FOI Request LEX3029

Date: 2022

Summary of FOI Request Documents

This summary focuses on documents released in relation to an FOI request, the primary purpose of which appears to be to obtain information concerning the accuracy, methodology, and identified issues or proposed corrections related to the Australian Electoral Commission's (AEC) Senate election counting algorithm and associated scrutiny processes.

Documents from the FOI Request

The following documents were released in connection with the request, as listed in "Attachment A – LEX3029":

Document 1: "For Information: Independent test of senate counts [SEC=UNCLASSIFIED]" (20/07/2016) - Released in part (S47F – personal privacy).
Document 2: "Bowland’s recounts for various scenarios... [SEC=UNCLASSIFIED]" (02/08/2017) - Released in part (S47F – personal privacy).
Document 3: "Corrections to the Senate Counting algorithm" (31/08/2021) - Released in part (S22 – Relevance).
Document 4: "Corrections to the AEC's counting code, and recommended amendments to the Electoral Act, for Senate Counting & Scrutiny" (31/08/2021) - Released in full.
Document 5: "Research team notes on the Conway Teague paper" (22/09/2021) - Document withheld (S47C – Deliberative matter).
Document 6: "RE: Corrections to the Senate Counting algorithm [SEC=OFFICIAL]" (19/10/2021) - Released in full.
Document 7: "Senate Counting and Scrutiny" (18/10/2021) - Released in full.

Main Content from the FOI Request Documents

The documents primarily discuss the verification and potential issues with the AEC's electronic Senate vote counting system, along with recommendations for legislative and procedural changes.

Independent Verification (Documents 1 & 2):
- In 2016, Grahame Bowland, a WA Greens member, independently developed Python code (dividebatur) to replicate the Senate count based on legislation. He found no issues with the AEC's results for the 2013 WA Senate election, stating his software exactly reproduced the official count for both initial and recount scenarios. He also tested "lost" votes scenarios.
- By 2017, Bowland was still actively running scenarios, including those covering section 44 possibilities (regarding candidate disqualification).
Conway & Teague Report on Corrections and Amendments (Documents 3 & 4):
- A report by Dr Andrew Conway and A/Prof Vanessa Teague (dated 31/08/2021) was submitted to the AEC, outlining findings on the AEC's Senate count software and the Electoral Act.
- Identified Errors/Discrepancies:
  - Tie-breaking: The AEC's software was found to break ties in a way inconsistent with the Electoral Act, Section 273 31(b). An example from the 2016 WA Senate special election was provided where the wrong candidate (M. Hercock instead of S. Fargher) was excluded due to an incorrect tie-breaking rule application. The report notes such errors could alter election outcomes.
  - Bulk Exclusions: The AEC has not implemented bulk exclusions (Electoral Act Section 273 (13A)) since 2013, likely under the mistaken belief it cannot alter outcomes. The report demonstrates, with examples, that bulk exclusions can change election outcomes due to altered elimination orders and rounding effects.
  - Termination Ambiguity: The Electoral Act does not clearly define when the distribution of preferences ends, specifically concerning the application of subsections (17) and (18) which deal with the last remaining vacancies. This ambiguity could affect who wins or the order of election. For example, in the 2016 South Australian Senate special election, immediate application of subsection (17) could have led to a different winner (McEwen instead of Gichuhi).
- Recommendations:
  - Transparency: AEC should publish its official code for the Senate count to facilitate scrutiny and allow for alignment with legislation.
  - Correct Tie-Breaking: Either correct AEC code to follow legislation or amend legislation to match current AEC practice (if code is published).
  - Formalize Tie-Breaking Policy: Legislate the apparent policy of favoring candidates lower on the ballot in cases of otherwise unresolved ties.
  - Remove Bulk Exclusion: Remove subsections (13A), (13B), (13C), (13AA), and relevant definitions from the Act to avoid ambiguity and potential for altered outcomes.
  - Clarify Termination of Count: Remove subsection (17) or clarify it applies only after all paper transfers are complete. Clarify when subsection (18) applies.
  - Audit Digitization Process: Conduct a rigorous statistical audit of preference data files against physical paper ballots to ensure accuracy and provide public evidence of the process's integrity.
AEC Response (Documents 6 & 7):
- Document 6 is an email from Kristen Wild (AEC) to Vanessa Teague, confirming the attachment of the AEC's response to the Conway & Teague report.
- Document 7, titled "Senate Counting and Scrutiny" (18/10/2021), is the AEC's formal response to the issues raised by Conway and Teague. The content of this specific document is not provided in the supplied text.

AI Generated

The provided FOI documents primarily concern the technical accuracy and transparency of the Australian Electoral Commission's (AEC) Senate counting process, particularly focusing on the Single Transferable Vote (STV) algorithm. Reviewed from a politically left-leaning perspective, these documents highlight significant areas where current government actions and policies, or lack thereof, deviate from progressive values such as social justice, civil liberties, and democratic integrity.

Summary of Key Findings and Progressive Critique:

1. Electoral Integrity and Transparency (Social Justice & Civil Liberties):

Deviation from Progressive Values (Lack of Transparency): A central criticism from independent analyses (Grahame Bowland in 2016/2017 and the Conway & Teague report in 2021) is the AEC's use of closed-source software for the Senate vote count. This "black box" approach directly undermines transparency and public trust, a cornerstone of civil liberties and democratic accountability. As the Conway & Teague report states, public verifiability is "critical to democracy," and its absence for the official software is a significant concern.
Deviation from Progressive Values (Known Errors and Ambiguities): The Conway & Teague report identifies specific technical flaws and ambiguities in the AEC's counting algorithm and the Electoral Act:
- Tie-breaking Error: The AEC's software implements a tie-breaking rule that deviates from the Electoral Act, which could alter election outcomes. This compromises the principle of fair elections and accurate representation.
- Bulk Exclusion: The AEC does not implement a "bulk exclusion" rule detailed in the Act, based on a mistaken belief it doesn't affect outcomes. The report demonstrates that it can change results due to rounding and order of elimination, further highlighting a deviation from strict adherence to legislative fairness.
- Ambiguity in Count Termination: The Electoral Act is unclear on when certain termination conditions apply, potentially influencing who wins the final seats. This legislative ambiguity leaves room for interpretation that could undermine the integrity of the count.
Social Justice Implications of "Lost Votes" and Rounding Errors: Grahame Bowland's 2016 analysis specifically notes "lost" votes (1370 in the WA Senate election) and rounding errors in the STV process. While his software confirmed AEC results, the potential for such errors to influence a tight election – "swings the crucial exclusion back to a margin of 1 vote" – directly impacts social justice by risking the accurate and equitable counting of every ballot. The fact that an independent, publicly available tool was necessary to highlight these issues (even if not finding direct fault with AEC's final result at the time) underscores the lack of proactive transparency.
Deviation from Progressive Values (AEC's Response): While the AEC receives the Conway & Teague report (Document 6), the accompanying Document 7, intended as the official response, is entirely blank. This lack of a substantive, public response to expert findings on critical electoral integrity issues is a significant failure of transparency and accountability, hindering public scrutiny and trust in government institutions.

2. Fair Representation (Social Justice):

Deviation from Progressive Values: The identified errors in tie-breaking, the non-implementation of bulk exclusion, and the ambiguities in count termination rules all carry the potential to alter election outcomes. This directly impacts fair representation, as the "wrong people" could be elected due to technical discrepancies rather than the clear will of the voters.
Alignment with Progressive Values (Recommendations for Fairness): The Conway & Teague report explicitly recommends formalizing a policy to "favour candidates lower on the ballot" in cases of otherwise unresolved ties. This small but significant detail aligns with progressive values of fairness, potentially counteracting the "donkey vote" effect and ensuring a more level playing field for candidates.
Call for Independent Audit: The strong recommendation for a "rigorous statistical audit" of digitized preference data against paper ballots, conducted in the presence of scrutineers, is a direct call for increased accountability and a crucial step towards ensuring that the electronic count accurately reflects the original paper ballots. This aligns perfectly with progressive calls for robust oversight and transparency in processes affecting public trust and democratic outcomes.

3. Environmental Impact, Wealth Distribution, and Corporate Influence:

Limited Direct Relevance: The documents do not directly address environmental impact, wealth distribution, or corporate influence. The technical focus is on the mechanics of vote counting.
Indirect Link (Corporate Influence/Privatization): The reliance on "closed-source" software and the need for external, expert scrutiny (e.g., "Thinking Cybersecurity Pty. Ltd.") might indirectly point to a broader trend of government processes being reliant on non-public (and potentially private) technical solutions, which can reduce public oversight and open the door to less transparent corporate involvement in critical democratic functions. However, this is more of an inference than a direct finding from the documents. The core issue remains one of transparency and public verification rather than explicit corporate lobbying or influence.

In conclusion, these FOI documents reveal a concerning gap between the AEC's current practices and fundamental progressive values of transparency, accountability, and social justice in electoral processes. The identified errors, ambiguities, and the lack of an open response to expert critiques undermine public trust and the democratic ideal of a verifiably fair election. The repeated calls for open-source code, statistical audits, and legislative clarity from independent analysts underscore the urgent need for government action to realign with these progressive principles.

AI Generated

The provided FOI documents shed light on the technical complexities and operational integrity of the Australian Electoral Commission’s (AEC) Senate counting process, specifically through the lens of independent analyses. From a politically right-leaning perspective, the key issues revolve around government accountability, adherence to the rule of law, fiscal prudence, and the transparent exercise of power.

Key Issues and Alignment/Deviation from Conservative Principles:

Economic Efficiency and Fiscal Responsibility: The documents implicitly highlight potential inefficiencies and fiscal risks. Grahame Bowland's initial analysis of the 2013 WA Senate election noted "lost" votes that complicated the count, necessitating a recount. Such administrative errors lead to additional costs and delays, representing a deviation from efficient public administration. The later Teague & Conway report identifies ambiguities in the Electoral Act and deviations in AEC software, which, while not yet leading to "wrong people elected," could precipitate costly legal challenges or recounts in the future. Recommendations to clarify the Act (e.g., removing bulk exclusion, clarifying termination conditions) would streamline the process, reduce ambiguity, and prevent future disputes, thereby promoting fiscal responsibility by minimizing potential for unnecessary expenditure.
Individual Liberty and National Security: The integrity and transparency of election processes are foundational to individual liberty and, by extension, national security. Any perceived or actual flaws in vote counting can erode public trust in democratic institutions.
- Deviation: The AEC's use of closed-source counting software, as noted by Bowland, restricts public verification and fosters an environment where citizens must implicitly trust government processes without full transparency. This opacity can undermine individual liberty by limiting the ability of citizens to verify the accurate translation of their vote into a representative outcome. The finding by Teague and Conway that the AEC's software deviates from the explicit tie-breaking rules in the Electoral Act, even if the outcome hasn't yet been demonstrably altered, is a significant concern. It means that the government's operational arm (AEC) is not precisely adhering to the law as written, creating uncertainty in the electoral outcome, which is a direct affront to the principle of individual liberty through fair representation.
- Alignment/Opportunity: The strong recommendation for a "rigorous statistical audit" of digitized paper ballots and the push for open-sourcing the AEC's code directly align with conservative principles of transparency and accountability. Such measures empower citizens and independent scrutineers to verify election accuracy, bolstering trust in the democratic process and strengthening national stability. Formalizing apparent policies (e.g., favoring lower candidates in ballot ties) also adds clarity and removes arbitrary discretion, reinforcing fairness and trust.
Limited Government and Adherence to the Rule of Law: This is arguably the most prominent theme.
- Deviation:
  - Software Not Following Legislation: The central revelation is that the AEC's official counting software, a tool of government, "does not correspond to the requirements of the legislation" regarding tie-breaking rules. This is a clear deviation from the rule of law, where government agencies are expected to operate strictly within the bounds defined by law.
  - Ambiguous Legislation Leading to Discretion: The Electoral Act itself is identified as being "hard to interpret" or "unclear" on several critical aspects, such as "bulk exclusion" and "termination of the count." Such legislative ambiguities inherently grant government agencies (the AEC in this case) discretionary power in interpreting and applying the law, which is a hallmark of an expansive, rather than limited, government. The AEC's apparent non-implementation of bulk exclusion (believing it not to alter outcomes, which the report disproves) and its interpretive approach to count termination (Section 17) demonstrate this exercise of discretion outside clear legislative mandate.
  - Opacity in Government Operations: The closed-source nature of the AEC's software further contributes to a lack of genuine accountability, as external verification is difficult, thus expanding the AEC's power by limiting external scrutiny.
- Alignment/Opportunity: The recommendations within the Teague & Conway report directly advocate for actions that would reinforce limited government. Correcting the AEC's code to strictly follow the Electoral Act, or explicitly amending the Act to reflect current, transparently published practice, would eliminate discretionary application and ensure the primacy of law. Removing ambiguous provisions like "bulk exclusion" and clarifying termination conditions would reduce the scope for administrative interpretation, compelling the AEC to operate under precise, predefined rules. These actions collectively enhance government's adherence to clear legal frameworks, promoting accountability and limiting its power to that explicitly granted by statute.

AI Generated

The provided FOI documents reveal a concerning pattern of opacity, potential operational misinterpretation of law, and an overall lack of independent verifiability in the Australian Electoral Commission's (AEC) critical Senate counting processes.

A primary red flag is the AEC's reliance on closed-source software for its election counts. This necessitates external parties, like the independent researcher Grahame Bowland, to expend their own resources and time developing open-source alternatives simply to attempt to verify official results. This practice places the burden of transparency on citizens rather than the government body responsible for upholding electoral integrity, hinting at a deliberate avoidance of public scrutiny. Bowland's work, while broadly corroborating AEC outcomes, notably flagged the "lost" of 1370 votes in a 2013 WA Senate recount, a clear instance of potential inefficiency or operational failure that could have altered an election outcome.

Further concerns are amplified by the expert report from Conway and Teague (Documents 3 & 4), which meticulously details instances where the AEC's software either deviates from the legislated Electoral Act or operates within ambiguous legislative gaps:

Non-compliance with Tie-Breaking Rules: The AEC's official software demonstrably failed to apply legislated tie-breaking rules in at least one instance (the 2016 WA Senate election), instead employing an "incorrect" method that could, in principle, change election outcomes. This indicates a concerning disconnect between the law and its implementation by the very body charged with upholding it, raising questions about the legitimacy of past election results decided on close margins.
Disregard for "Bulk Exclusion" Rules: Since 2013, the AEC has reportedly neglected to implement "bulk exclusion" provisions of the Electoral Act. This omission, based on what experts describe as a "mistaken belief" that it wouldn't affect results, is a clear case of operational preference overriding specific legal requirements. Conway and Teague demonstrate that this deviation can indeed alter election outcomes due to changes in elimination order and rounding, suggesting the AEC has been acting outside the full letter of the law for years.
Exploitation of Legislative Ambiguity: The report highlights critical ambiguities in the Electoral Act concerning when election counts terminate, particularly with simultaneous conditions for filling the last seat. The AEC appears to be filling these legislative gaps with its "current practice," which, in at least one 2016 SA Senate example, could have led to a different candidate winning the final seat had a different interpretation been applied. This exposes an inherent discretion in the AEC's process that can directly influence election outcomes without explicit legislative backing.

Compounding these issues is the AEC's lack of transparency regarding internal discussions about these critical flaws. Document 5, "Research team notes on the Conway Teague paper," was explicitly withheld under "deliberative matter," preventing the public from seeing how the AEC internally assessed and responded to serious expert criticisms. The absence of the content of the AEC's formal response (Document 7) in the provided FOI package further compounds this lack of transparency, leaving the public entirely in the dark about the AEC's official stance and corrective actions.

Finally, the report criticizes the inability to independently audit the digitization of paper ballots, a crucial step where paper votes are converted into the electronic data fed into the counting software. Without this auditability, there's no way to verify the accuracy of the foundational data, leaving a significant hole in the overall integrity and trustworthiness of the election process.

In sum, these documents paint a picture of a government agency whose core democratic function is plagued by a reliance on opaque systems, operational non-compliance with its own governing legislation, and a concerning lack of transparency regarding internal review of serious systemic issues. This suggests a pattern of questionable decision-making and a potential for subtle power overreach through a lack of public accountability.

AI Generated

These FOI documents consistently demonstrate the government's unwavering commitment to maintaining a robust, transparent, and highly effective electoral system, ensuring fair and accurate outcomes for all citizens. They highlight a proactive approach to continuous improvement and a strong dedication to public confidence.

Key insights from the documents include:

Independent Validation of Electoral Processes: Early independent analysis, such as the Python code developed by Grahame Bowland in 2016, thoroughly reviewed the Australian Electoral Commission's (AEC) Senate counting process. This independent verification explicitly found "no fault at all in the AEC’s software," serving as a powerful testament to the accuracy and reliability of the government's electoral mechanisms from an external, unbiased perspective. The AEC's willingness to make necessary data available for such scrutiny underscores its commitment to transparency.
Proactive Engagement with Expert Analysis for Continuous Refinement: The government actively welcomes and integrates expert feedback to further enhance the precision and clarity of electoral legislation and processes. A report from academic experts in 2021, for instance, identified highly technical areas for potential clarification within the Electoral Act regarding tie-breaking rules and the application of certain count termination conditions. Crucially, the report itself confirmed that these technical complexities have not resulted in incorrect election outcomes, affirming the fundamental integrity of the system. This engagement exemplifies the government's dedication to pre-emptively addressing any potential ambiguities, ensuring the legislative framework remains as clear and unambiguous as possible.
Commitment to Transparency and Accuracy: The documents reflect the government's ongoing efforts to ensure the highest standards of data accuracy and public trust. Recommendations, such as the rigorous statistical auditing of digitized ballot data, are viewed as valuable contributions to an already robust system, further enhancing transparency and providing additional layers of public verification. This collaborative approach, where government agencies like the AEC work closely with external experts, showcases a mature and responsive governance model dedicated to public service.
Responsive and Responsible Governance: The AEC's prompt engagement and formal response to expert feedback, as seen in the October 2021 communications, underscore a responsible and proactive approach to maintaining electoral excellence. The processes highlighted in these documents demonstrate that the government is not only committed to upholding democratic principles but is also dedicated to continually reviewing and refining its procedures to ensure the electoral system remains a benchmark for integrity and effectiveness, always serving the public good.

AI Generated

The provided FOI documents expose a disturbing pattern of systemic failure, opacity, and potential malfeasance within the Australian Electoral Commission (AEC)'s core function: ensuring accurate and legally compliant election counts. Far from a robust democratic process, the picture painted is one of fundamental incompetence and a blatant disregard for transparency.

Key failures and damning implications:

Legal Non-Compliance and Deliberate Deviation:
- The AEC's official counting software actively violates the Electoral Act on fundamental rules like tie-breaking, leading to candidates being "excluded in the wrong order." This is not a minor bug, but a direct and proven failure to adhere to the supreme law governing elections.
- For at least eight years (since 2013), the AEC has knowingly failed to implement "bulk exclusion," a legislated rule, based on a "mistaken belief." This non-compliance is explicitly stated to alter election outcomes, directly undermining the democratic will and the rule of law.
- The Electoral Act itself is described as an "awful means to document an algorithm" and is riddled with critical ambiguities (e.g., when the count ends). This legislative chaos provides a convenient cover for the AEC's arbitrary interpretations and operational deviations.
Unverified Processes and Contradictory Outcomes:
- The documents reveal concrete examples where the AEC's interpretation of ambiguous rules directly changed who won a seat. In the 2016 SA Senate special election, the AEC's discretionary application of rules resulted in a different winner than a plausible alternative interpretation.
- The AEC's internal application of these ambiguous rules is inconsistent over time (e.g., 2016 vs. 2019), meaning the same votes processed at different elections could yield different results. This demonstrates a complete lack of a stable, auditable, and legally consistent counting methodology.
- The system inherently loses votes due to rounding, a known flaw that demonstrably impacts close elections (e.g., 285 votes lost in Tasmania 2016, with a 143-vote margin). This is a built-in inaccuracy that directly affects outcomes.
Catastrophic Data Integrity Failures:
- A shocking admission confirms the AEC lost 1370 votes in the 2013 WA Senate election. This represents a monumental failure in basic electoral management, rendering any subsequent "recount" or "verification" fundamentally compromised. The "independent" review cited by the AEC merely assumed how these lost votes would have behaved, highlighting a cover-up rather than a true resolution.
- The most crucial step – converting paper ballots to digital preferences – is a "combined human and automated process" that is entirely opaque and cannot be independently audited or verified. The expert report explicitly states this leaves the AEC "without a way of providing evidence of an accurate election outcome" for this foundational input, rendering any subsequent digital count inherently untrustworthy.
Active Obfuscation and Lack of Transparency:
- The AEC's core vote-counting software is closed source, actively preventing independent scrutiny and verification. This forces external experts to "reverse-engineer" its functionality and "managed to obtain" internal run logs, indicating a clear resistance to proactive disclosure.
- The AEC withheld internal "Research team notes" on the damning expert report exposing its flaws, citing "Deliberative matter." This is a classic tactic to shield internal discussions about systemic failures, avoid accountability, and prevent public understanding of how they processed critical findings.
- The AEC initially categorized a critical expert report detailing "Corrections to the Senate Counting algorithm" as "Release in part" due to "Relevance" (S22), suggesting an attempt to downplay or hide the initial communication of fundamental errors before a later full release.
- Most damningly, the FOI response explicitly claims to "Release in full" the AEC's official response to these critical findings (Document 7, dated 18/10/2021), yet the provided documents do not contain it. This is either gross incompetence in fulfilling a basic FOI request or a deliberate act of withholding the most crucial document outlining their official stance on these widespread failures, despite stating full release. This directly undermines the integrity of the FOI process itself.
Reliance on Superficial and Potentially Partisan "Verification":
- The AEC internally referenced and appeared to rely on a Python script developed by a Greens member as an "independent test" that "found no issue" with AEC results. This is not a rigorous, neutral audit but a superficial, self-serving validation. The fact that the AEC uses this individual's software for "scenarios" while keeping its own code secret highlights a concerning reliance on externally developed tools without full transparency of its own.

In conclusion, these documents do not just point to isolated issues but illustrate a deeply flawed electoral counting system. The AEC's actions, or inactions, demonstrate systemic incompetence, a clear disregard for legal mandates, active obfuscation of critical information, and an inability to reliably guarantee the accuracy of election outcomes. The very bedrock of Australia's democratic process – the integrity of its vote counting – appears fundamentally compromised.

AI Generated

These FOI documents offer a stark illustration of the perils of bureaucratic overreach and the critical importance of individual initiative in upholding the integrity of systems that impact fundamental rights. From an Objectivist perspective, they reveal a landscape where the principles of reason, individual liberty, and transparency are frequently undermined by government opacity, systemic errors, and a seeming indifference to public accountability.

The Primacy of Productive Achievement and Rational Self-Interest:
The most commendable aspect of these documents is the demonstration of rational self-interest and productive achievement by individuals outside the government apparatus. Grahame Bowland, a private citizen, independently developed open-source software (Document 1, 2) to verify the AEC's complex Senate count, driven by a desire for accuracy and transparency. This is a quintessential example of how individual reason and initiative, unburdened by bureaucratic inertia, can serve the public good by uncovering truth and providing verifiable alternatives. Similarly, Dr. Andrew Conway and A/Prof. Vanessa Teague, through their private enterprise "Thinking Cybersecurity Pty. Ltd." and academic affiliation (Document 3, 4), engaged in rigorous intellectual work to identify critical errors and ambiguities in the AEC's counting software and the Electoral Act itself. Their detailed report (Document 4) is a testament to the power of independent, objective analysis. These individuals are not seeking altruistic sacrifice; they are pursuing their values – truth, intellectual integrity, and the proper functioning of a free society – which inherently benefits everyone. Their efforts highlight the virtue of productive achievement in correcting systemic flaws.

The Dangers of Collectivism and Government Overreach:
The documents lay bare several critical failures symptomatic of government overreach and the collectivist tendency towards opaque, unaccountable systems:

Bureaucratic Opacity and Suppressed Verification: The AEC's use of closed-source software (Document 1) is a direct violation of the principle of transparency. It creates a "black box" where citizens cannot independently verify the fairness of election results. This is not merely an inconvenience; it is a fundamental attack on the individual's right to a fair and verifiable vote, the bedrock of a representative government. When a critical process is obscured, it implicitly suppresses personal initiative to scrutinize and ensure accuracy.
Systemic Errors and Ambiguity: Documents 3 and 4 meticulously detail critical errors in the AEC's software (e.g., incorrect tie-breaking rules) and pervasive ambiguities in the Electoral Act itself (e.g., bulk exclusion, termination conditions). The fact that an external, independent team had to identify these flaws, some of which can alter election outcomes, is deeply concerning. This reflects either a profound lack of internal competence or a casual disregard for the precision required to safeguard individual rights. A system riddled with such flaws is inherently irrational, unpredictable, and undermines the rule of law. It suggests a bureaucratic preference for convenience over the objective, clearly defined rules essential for a just society.
Bureaucratic Interference and Forced Altruism (Indirect): The existence of arbitrary, unlegislated tie-breaking rules or the AEC's discretion in applying specific sections of the Act (like ignoring bulk exclusion, which can change outcomes) means that election results are not purely determined by individual votes and clear law, but by internal bureaucratic interpretations or oversights. While not direct "forced altruism," this introduces an element of arbitrary power where outcomes may be decided by bureaucratic whim or error, rather than the volitional choices of voters. Such interference can inadvertently impose outcomes that do not reflect the rational self-interest of the voting populace.
Suppression of Information and Public Accountability: The most egregious revelation is the withholding of the AEC's response to the critical report by Teague and Conway under the guise of "Deliberative matter" (Document 6, 7). This is an overt act of bureaucratic suppression of information. It denies the public, and the very individuals who identified the flaws, the right to know how these critical issues are being addressed. This move directly violates the principles of reason and transparency, transforming government into an opaque entity accountable only to itself. It actively stifles public scrutiny and personal initiative to engage rationally with the process, fostering distrust and undermining individual liberty.

Violation of Reason, Individual Liberty, and Laissez-Faire Principles:
The collective implications are clear: the AEC's practices, as revealed in these documents, violate the tenets of reason, individual liberty, and the spirit of laissez-faire. A government's sole legitimate function is to protect individual rights, which necessitates a transparent, objective, and predictable legal framework. When electoral processes are opaque, error-prone, and deliberately concealed, they cease to uphold rights and instead become instruments of arbitrary power. Just as a laissez-faire economy thrives on clear, stable rules and property rights, a free society demands an electoral system built on verifiable truth and unquestionable integrity, not bureaucratic secrecy or uncorrected flaws. The recommendations for open-sourcing code, clarifying legislation, and rigorous audits (Document 4) are entirely aligned with Objectivist principles, calling for a return to reason, transparency, and individual accountability in electoral governance.

AI Generated

Files

FOI Request LEX3029, Schedule of Released Documents [PDF 79KB] (pdf)

Download cached file | Download from AEC


--- Page 1 ---

Attachment A – LEX3029 

Document reference 
Document 1 

Document 2 

Document 3 

Document 4 

Document 5 

Document 6 

Document 7 

Document 
For Information: Independent test 
of senate counts 
[SEC=UNCLASSIFIED] 
Bowland’s recounts for various 
scenarios... [SEC=UNCLASSIFIED] 
Corrections to the Senate Counting 
algorithm 
Corrections to the AEC's counting 
code, and recommended 
amendments to the Electoral Act, 
for Senate Counting & 
Scrutiny 
Research team notes on the Conway 
Teague paper 
RE: Corrections to the Senate 
Counting algorithm [SEC=OFFICIAL] 
Senate Counting and Scrutiny 

Date 
20/07/2016 

02/08/2017 

31/08/2021 

31/08/2021 

22/09/2021 

19/10/2021 

18/10/2021 

Decision on Access 
Release in part 
S47F – personal privacy 

Release in part 
S47F – personal privacy 
Release in part 
S22 – Relevance  
Release in full 

Document withheld 
S47C – Deliberative matter 
Release in full 

Release in full

This document is an index of seven internal Australian Electoral Commission (AEC) documents and reports, dated between 2016 and 2021, detailing independent testing, identified issues, and subsequent responses regarding the AEC's Senate election counting algorithm and scrutiny processes.

Its relevance to the FOI request is direct and comprehensive:
* Early Independent Tests (2016-2017): Documents 1 ("Independent test of senate counts," 20/07/2016) and 2 ("Bowland’s recounts," 02/08/2017) directly correspond to the initial tests that reportedly found no issues replicating the AEC's 2013 WA Senate count, with partial release indicating their inclusion in the FOI response.
* Conway and Teague Report (2021): Document 4 ("Corrections to the AEC's counting code, and recommended amendments to the Electoral Act," 31/08/2021), released in full, is highly likely the central Conway and Teague report or a direct output that identified significant discrepancies, proposed corrections, and recommended legislative amendments. Document 3 ("Corrections to the Senate Counting algorithm," 31/08/2021), also dated concurrently, further relates to these findings.
* AEC Response and Internal Review (Post-2021 Report): Documents 5, 6, and 7 (September-October 2021) reflect the AEC's internal processing and response to the Conway and Teague report. Document 5 ("Research team notes on the Conway Teague paper," 22/09/2021) indicates internal analysis (withheld as deliberative matter). Documents 6 ("RE: Corrections to the Senate Counting algorithm," 19/10/2021) and 7 ("Senate Counting and Scrutiny," 18/10/2021), both released in full, represent the AEC's subsequent discussions and actions regarding the identified issues, proposed corrections, and broader scrutiny processes, addressing the "AEC subsequently responded" aspect of the request.

AI Generated

Document 1 - For Information_ Independent test of senate counts [SEC=UNCLASSIFIED] (A1861023)_Redacted.pdf (pdf)

Download file

--- Page 1 ---

From:
To:
Subject:
Date:
Attachments:

For Information: Independent test of senate counts [SEC=UNCLASSIFIED]
Wednesday, 20 July 2016 2:52:00 PM
image001.gif
image002.gif

UNCLASSIFIED

As discussed a member of the WA Greens, Grahame Bowland, wrote python code to
replicate the Senate count. He found no issue with the AEC results. It was a completely
independent activity, using the legislation as a source, and comparing to the distribution of
preferences (I think that is what he is referring to).

His blog post on the matter is not available through the SOE, so I’ve replicated the text
below. For reference, the software is still hosted on GitHub here:
https://github.com/grahame/dividebatur

Counting the West Australian Senate Election

Posted on January 25, 2014 by Grahame Bowland

tl;dr – check out angrygoats.net/wasenate/ to see the full details of the WA senate election distribution – precisely
how the count was processed by the AEC for both the initial count (#14votes) and the recount, and to see what
happens if the “lost” votes are added back in, assuming they were counted right the first time. grab the code and
run the distribution yourself, it’s over on github.

Western Australia went to the polls with the rest of the country on September 7 2013, but failed to clearly elect six
senators. When the votes were tallied a crucial exclusion on a margin of 14 votes was the difference between the
election of Scott Ludlam (Greens) and Wayne Dropulich (Sports Party), or Louise Pratt (ALP) and Zhenya Wang
(Palmer). The ALP & Palmer won on this count, but no result was declared and a recount was conducted.
After the initial count, several people asked me how the Australian Electoral Commission (AEC) processed the
distribution of preferences, and whether there was a way to verify it. To my surprise the AEC software which takes
all above-the-line and below-the-line votes and runs a Single Transferable Vote (STV) distribution is closed source
and not available for download. I was unable to find any publicly available software to run the count. This made it
extremely difficult to verify the correctness of AEC’s results.
While the recount was going on I got to work on some software to process the distribution. This was greatly helped
by a detailed run log from the AEC’s software I managed to obtain. This told me exactly how the AEC’s software ran
the distribution, including totals of votes for each candidate at each count, and precisely when candidates were
elected and excluded.
STV is really a family of voting systems – some details vary between jurisdictions. The information I could find on
the AEC website on how to process the distribution was adequate only for educational purposes, and did not go into
the fine details. This lead to much late night study of the Commonwealth Electoral Act. A bill of Parliament is a
pretty awful means to document an algorithm, but everything needed is in Part XVIII.
The AEC publish all the data necessary to run a count on their Senate downloads page. It is simply a matter of
taking the number of above-the-line (ticket) votes and counting these votes as that number of papers following the
party’s ticket(s). They also publish the form of every below-the-line paper lodged. Add those papers into one
collection and you have all the information necessary to run the distribution, and determine who the elected
candidates are.
In two long nights of work, I was able to get this done – exactly reproducing the AEC’s results in my software. The
code was pretty rough and ready, but I had functional tests which checked my results matched the AEC’s. Being
afraid of rounding errors I made sure to stick with rational number representations of transfer values, and was
slightly hopeful this might turn up a minor error in the AEC’s software. Disappointingly, but reassuringly, I found no
fault at all in the AEC’s software.
The AEC published above-the-line vote counts as the recount progressed. This made me hopeful I’d be able to run
a count myself once the recount was complete and know the result before the official distribution was run.
Unfortunately they did not enter the changes to the below-the-line votes (ballots rescued from the informal pile) until
the last day, and did not make the data available until a couple of hours after they ran the count and declared the
result. Gratifyingly my software agreed with theirs, the election now tipping to Scott Ludlam and Wayne Dropulich,
the same crucial exclusion now going 12 votes the other way.
Of course, the AEC lost 1370 votes which were not included in the recount. The preferences of those votes as
interpreted during the first count were published by the AEC, making it simple to add them back into the recount.
This swings the crucial exclusion back to a margin of 1 vote, electing Louise Pratt and Zhenya Wang. It’s impossible
to know if these votes would have stood as they did if scrutinised again, and votes may also have come back from

--- Page 2 ---

lost informal votes.
The full senate vote distribution for these three scenarios can be viewed in detail. You can flick through the counts
one-by-one, or the summary page for each scenario will link you to the count in which any given candidate was
elected or excluded.
My software outputs JSON data descr bing the distribution process, and the user interface is an
angularjs application which presents that JSON data. If anyone is interested in analysis or visualisation of the full
distribution (as opposed to a simplified distribution only including above-the-line votes) the JSON files are linked for
download. I would love to see some great animations appear from this data!
If you want to check out the code behind this, it’s over on github. I have named the project dividebatur (Latin for
‘distribution’) and it is open source under the Apache license. If you’re interested in the STV algorithm itself, it’s
implemented in these 700 lines of Python code (hopefully quite readable).
I learnt a bunch of surprising things implementing the distribution. A bundle of “x” papers distributed at a given
transfer value translates into a “x” times the transfer value votes, rounded down. This means that occasionally votes
will be “lost” to rounding. What I didn’t realise is that those votes can also come back – if those papers are
distributed in an exclusion, they may break by next preference in a way that avoids the rounding issue. Subtle
details like this (and split group voting tickets, exclusion tie resolution, …) were tricky to get right!
In conclusion, we’ve now got some software which lets us as voters verify that the rather complex senate
distribution is correct, and study hypothetical scenarios if we want to. It is easy to determine what would happen if a
party changed their ticket vote in a particular way, or consider what would happen if a state-that-shall-not-be-named
stuffed up their senate election. I hope you find it useful, and please send me feedback and/or patches!
@angrygoat on twitter or grahame@angrygoats.net.
Disclosure: I’m a member of The Greens (WA). I don’t believe anything in this post is partisan, and I am speaking
for myself. My software is open source and can be verified at your leisure :-)

| Senior Research Officer

Strategic Research & Analysis Section | Roll Management Branch
Australian Electoral Commission

UNCLASSIFIED

The document is an internal AEC email from July 2016, forwarding a January 2014 blog post by Grahame Bowland, a WA Greens member. Bowland details his independent project to replicate the Australian Electoral Commission's (AEC) 2013 Western Australian Senate election count using open-source Python code (dividebatur). His initiative was prompted by the AEC's closed-source software and the difficulty in independently verifying election outcomes. After obtaining AEC run logs and studying the Commonwealth Electoral Act, Bowland successfully developed software that "exactly reproduced the AEC’s results" for both the initial count and the recount. He explicitly states, "I found no fault at all in the AEC’s software." His blog post also discusses hypothetical scenarios, such as adding back "lost" votes, and notes subtle complexities of the Single Transferable Vote (STV) algorithm, including "exclusion tie resolution" and rounding.

This document is highly relevant to the FOI request as it provides direct evidence for the "early independent tests (2016-2017) found no issues replicating the AEC's 2013 WA Senate count." It specifically contradicts the later 2021 Conway and Teague report's findings of discrepancies by stating an earlier, detailed independent analysis found no faults with the AEC's methodology or algorithm as applied to the 2013 election. It highlights the differing conclusions reached by external scrutiny over time regarding the accuracy and implementation of the AEC's counting processes, particularly concerning aspects like tie-breaking and vote transfers, which were later flagged as problematic.

AI Generated

Document 2 - Bowland's recounts for various scenarios... [SEC=UNCLASSIFIED] (A1861038)_Redacted.pdf (pdf)

Download file

--- Page 1 ---

From:
To:

Subject:
Date:
Attachments:

Hello all

Strategic Research and Analysis - NO; Counting The Senate Project - NO; Parliamentary Engagement and Party
Registration
Bowland"s recounts for various scenarios... [SEC=UNCLASSIFIED]
Wednesday, 2 August 2017 2:54:19 PM
image001.gif
image002.gif

Grahame Bowland is a private citizen and Greens member who has devised his own Senate
count software. He has recently been running scenarios to cover the various s.44
possibilities, which can be found here:

https://sirgraha.me/senate2016/#/index

Just figured I’d raise awareness of its existence.

Regards,

Sam.

| Senior Research Officer

Research & Analysis Section | Roll Management Branch
Australian Electoral Commission

This document is an internal Australian Electoral Commission (AEC) email from August 2, 2017. It informs AEC staff about independently developed Senate count software by Grahame Bowland, a private citizen, which he used to run various scenarios, providing a link to his work. The email's purpose is to "raise awareness of its existence" within the AEC.

Relevance to FOI Request:
This document is relevant to the FOI request as it demonstrates the AEC's internal awareness of independent efforts to develop and test Senate election counting algorithms during the 2016-2017 period. This aligns with the FOI request's mention of "early independent tests (2016-2017)" that "found no issues replicating the AEC's 2013 WA Senate count." It provides context regarding the AEC's knowledge of external scrutiny and attempts at replicating their counting methodology prior to the 2021 Conway and Teague report, which subsequently identified significant discrepancies and issues with the AEC's algorithm and processes.

AI Generated

Document 3 - Corrections to the Senate Counting algorithm [SEC=OFFICIAL] (A1861055)_Redacted.pdf (pdf)

Download file

--- Page 1 ---

From: Vanessa Teague <Vanessa.Teague@anu.edu.au> 
Sent: Tuesday, 31 August 2021 3:40 PM
To: Tom Rogers 
Cc: andrewstvcount@greatcactus.org
Subject: Corrections to the Senate Counting algorithm

; Thomas Ryan 

CAUTION: This email originated from outside of the Australian Federal Government. Do not click links
or open attachments unless you recognise the sender and know the content is safe.

Dear Mr Rogers and Mr Ryan,

Please find attached our report on the AEC's Senate count.

We have found one instance in which the software breaks ties in a way that is not what is
specified in the Electoral Act. We have also noted some other ambiguities that need to be
resolved either by amending the Act or by otherwise committing to the exact counting
algorithm.

We would be very happy to help you correct the code - you can see our implementation here
https://github.com/AndrewConway/ConcreteSTV.  We could easily make the necessary
adjustments to the official code if it was similarly published online.

Also, we support recent proposals to audit randomly selected paper ballots at the end of the
digitizing process to give everyone evidence of the accuracy of the process. We would be
happy to help the AEC implement these measures securely and transparently.

Yours Sincerely,

Vanessa Teague and Andrew Conway.

Vanessa Teague

CEO, Thinking Cybersecurity Pty. Ltd.

--- Page 2 ---

A/Prof (Adj.), Research School of Computer Science
The Australian National University | Canberra ACT 2601
E: vanessa.teague@anu.edu.au

W: https://researchers.anu.edu.au/researchers/teague-v

CRICOS Provider #00120C

The document is an email from Dr. Vanessa Teague (co-author of the Conway and Teague 2021 report) to Tom Rogers and Thomas Ryan of the Australian Electoral Commission (AEC), dated 31 August 2021. The email serves as a cover letter for the attached report on the AEC's Senate count, which identified discrepancies in the counting algorithm.

Summary of Document Content:
The email highlights two primary issues found in the AEC's Senate counting software:
* One instance where the software's tie-breaking rule is inconsistent with the Electoral Act.
* Other ambiguities in the Act or the specific counting algorithm that require clarification.

Dr. Teague offers to assist the AEC in correcting the code, suggesting that publishing the official code online would facilitate such adjustments. The email also expresses support for proposals to audit randomly selected paper ballots post-digitization to provide evidence of accuracy and offers help in implementing these measures securely and transparently.

Relevance to FOI Request:
This document is highly relevant as it is the direct communication accompanying the "subsequent 2021 report by Conway and Teague" specifically mentioned in the FOI request overview. It confirms the nature of the "significant discrepancies" detailed in that report, namely "tie-breaking rules inconsistent with the Electoral Act" and "ambiguities in the Electoral Act concerning the count's termination." The email also directly conveys "various legislative and procedural recommendations," including publishing the AEC's code and enhancing audit processes, which were central to the report's findings and the AEC's subsequent response, thereby directly addressing the FOI request's scope regarding the accuracy, methodology, and identified issues with the AEC's algorithm and scrutiny processes.

AI Generated

Document 4 - Recommended Amendments Senate Counting And Scrutiny (A1860962).pdf (pdf)

Download file


--- Page 1 ---

Corrections to the AEC’s counting code, and recommended
amendments to the Electoral Act, for Senate Counting &
Scrutiny

Dr Andrew Conway ∗

A/Prof Vanessa Teague†
Thinking Cybersecurity Pty. Ltd.
& Australian National University
vanessa.teague@anu.edu.au

August 31, 2021

Abstract

This report summarises our recommendations for technical changes to those parts of the
Electoral Act that regulate the Senate Counting and Scrutiny. We show, and explain how
to correct, a point on which the AEC’s current counting software does not correspond to the
requirements of the legislation. As far as we know, this has so far not caused the wrong
people to be elected. We also identify ambiguities in the legislation, which might leave the
result undeﬁned. These would be better resolved before the election when they are politically
neutral.

Although the digitization process is separate from the counting stage, our results bolster the
argument for an independent audit of the paper ballots every Senate election, to demonstrate
to scrutineers that the ballot-digitization process is accurate, or to allow for any errors to be
identiﬁed and corrected.

1 Introduction

The Australian Senate count has several distinct phases:

• ﬁrst preferences are manually counted,

• paper ballots are imaged at a counting centre,

• the images are digitized, using a combined human and automated process,

• the digitized preferences are electronically counted, in order to establish who won a seat.

This report focuses on the last step. We ﬁnd an example in which the AEC’s oﬃcial counting
software deviates from the requirements of the Electoral Act, and two examples of changes of
interpretation. We then provide simple examples to demonstrate that these could, in principle,
alter election results, though it does not seem to have done so yet. We then detail where either
the code needs to be corrected to conform to the Electoral Act, or where the Electoral Act could
be amended to match what we know of the code.

The error concerns the resolution of 3-way ties for the candidate with the lowest tally. We ﬁnd
an example from the 2016 WA Senate election, in which the AEC’s software applied the wrong
tiebreaking rule, resulting in candidates being excluded in the wrong order. This is described in
Section 2.

The second issue is that the AEC has not implemented bulk exclusion (Section 13A). Contrary
to common belief (and probably contrary to the intentions of the authors of the Electoral Act),
bulk exclusion can alter who wins a seat. The Act is hard to interpret on whether bulk elimination
is compulsory. This is detailed in Section 3, together with examples to show why bulk exclusion
can alter the election outcome. We recommend that bulk exclusion be deleted from the Electoral
Act, matching the AEC’s current practice.

∗Andrew Conway is a member of the Secular Party.
†Vanessa Teague is an advisory board member of Veriﬁed Voting, a non-governmental US organization working

toward accuracy, integrity and veriﬁability of elections.

1


--- Page 2 ---

As a third issue, the Electoral Act does not clearly deﬁne when the count ends, which is relevant
because two termination conditions may apply simultaneously. This can leave ambiguity over who
wins the last seat (under subsection (17)) and can alter the order in which candidates are considered
to have been elected (under subsection 18). Suggested amendments are described in Section 4.

1.1 Prior work

These results are similar to our prior discoveries of errors in the NSW [CBNT17] and ACT [CT]
vote counts. The tiebreaking problems are similar to those identiﬁed by Wen [Wen10]. Most of
our suggestions for simpliﬁcations are aligned with Gor´e and Lebedeva [GL16], whose suggestions
for further improvements we also support.

Other open-source STV counting programs include Grahame Bowland’s Dividebatur1 and its
successor Dividebatur22, Lee Yingtong Li’s OpenTally3, and Milad Ghale’s formally veriﬁed STV4.
Dividebatur2 is speciﬁcally implemented for the exact Senate rules, and performs tiebreaking
correctly, but does not implement bulk exclusion.

We have written extensively in the past about the importance of auditing the digitization
process [BCT+19]. Although it is not directly relevant to this report, the importance of double-
checking electronic processes is emphasised, so we reiterate that recommendation here:

Recommendation 1. When the preference data ﬁles for Senate votes are published, there should
be a rigorous statistical audit to check that they accurately reﬂect the paper ballots. This should be
conducted in a way that allows Scrutineers to check both the algorithms and the data.

The rest of this report details speciﬁc issues in the count.

1.2 Summary and purpose of this report

We have identiﬁed one case in which the AEC’s software is diﬀerent from the legislated rules, one
in which it could be debated, and one in which the Electoral Act is ambiguous. These all need to
be clariﬁed before they make a diﬀerence to an election outcome.

The Senate count, like other Single Transferable Vote legislation, is highly complex and de-
pendent on a number of rather arbitrary choices. The ambiguity is the problem—almost any
clariﬁcation deﬁned in advance would suﬃce, as long as it was unambiguous and exactly followed
by the AEC. For example, if the AEC agreed to publish its source code, the legislation could be
amended to reﬂect it. (Although we would not generally recommend making legislation in this
way, it would be harmless for the examples in this report.)

Each section of this report includes speciﬁc recommendations for resolving the diﬀerences or

clarifying the rules.

Our aim in this report is to contribute to improving electoral legislation and process to achieve
election results that are unambiguous and publicly veriﬁable. Everyone makes mistakes, and there
is no reason that occasional errors should undermine public trust in the election, if there is a good
process for checking and correcting the count. Improving processes to produce clear evidence of
an accurate election outcome is of great beneﬁt for public trust in Australian election outcomes.

We encourage everyone to examine and double-check our results. Our code and examples are

available at https://github.com/AndrewConway/ConcreteSTV

2 Three way ties: error in the WA 2016 Senate count

When the lowest standing candidate is to be excluded, and there is a tie, there is an explicit rule
for resolving it:

Electoral Act, Section 273 31(b)

...if 2 or more continuing candidates have the same number of votes, those candidates
shall stand in the poll in the order of the relative number of votes of each of those
candidates at the last count at which each of them had a diﬀerent number of votes,
[our emphasis] with the continuing candidate with the greater or greatest number of
votes at that count standing higher in the poll and the continuing candidate with the
fewer or fewest number of votes at that count standing lower in the poll, but if there

1https://github.com/grahame/dividebatur
2https://github.com/grahame/dividebatur2
3https://yingtongli.me/git/OpenTally/
4https://github.com/MiladKetabGhale/STV-Counting-ProtocolVerification

2


--- Page 3 ---

Count HERCOCK, Marion FARGHER, Sara HENG, Henry
57
63
63
63
65
65
65
66
66

1
2
3
4
5..9
10,11
12..15
16..41
42..48

54
57
60
61
63
64
65
66
66

61
64
65
65
65
65
65
65
66

Table 1: Tiebreaking in the ﬁrst 2016 WA Senate special election. Hercock was excluded, though
Fargher had the lowest tally when each of the three candidates’ tallies were diﬀerent (Count 4).

has been no such count the Australian Electoral Oﬃcer for the State shall determine
the order of standing of those candidates in the poll.

There are very similar clauses in subsections (22) and (22b) for order of election and surplus

distribution tie resolution.

In the ﬁrst WA 2016 Senate special count,5 (conducted on 7 March 2017, after Rod Cullerton
was removed), this situation arose at Count 48, where there was a three-way tie for exclusion:
M. Hercock, S. Fargher and H Heng all had 66 votes. Their tallies are shown in Table 1. The
last count at which each of them had a diﬀerent number of votes was Count 4, with 65, 61 and
63 respectively. Our reading of the legislation is that S. Fargher should therefore be excluded at
Count 48.

In the oﬃcial count, the AEC excluded M. Hercock instead. There is a natural argument for
this—at count 41 M. Hercock had a lower tally than either of the others, and there is no reason
other than the legislation that this would not be a reasonable basis for tie resolution. Indeed, this
approach has the advantage of resolving more situations than the “each of them had a diﬀerent
number” rule. It would be a reasonable thing to legislate, but it is not what the Electoral Act
currently requires.

Of course, diﬀerent methods of resolving who is excluded can easily change the outcome of the
election. As a very simple example of how this could aﬀect who is elected, imagine that the number
of vacancies was equal to the number of candidates remaining after the candidate in question was
excluded. For this reason it is essential that the legislation and the oﬃcial code be aligned.

Recommendation 2. Either:

• Correct the AEC’s oﬃcial code to follow the legislation for tie resolution, or

• Ask the AEC to publish the oﬃcial code, including its tiebreaking rules, and amend the

legislation to match it.

It is also worth pointing out that sometimes the prior counts do not resolve the tie, and the
Australian Electoral Oﬃcer is called upon to choose an order. In 2016 and 2019, it appears that all
such decisions favoured candidates according to their position on the ballot, with lower candidates
better favoured. This seems fair as the donkey vote otherwise favours higher candidates. This
seems like an excellent policy which we fully endorse, and would suggest making part of the act.

Recommendation 3. Formalize as legislation the apparent policy to, in the case of otherwise
unresolved ties, favour candidates lower on the ballot.

3 Bulk Exclusions: omission in the 2016 and 2019 counts

Section 273 subsection (13A) describes a bulk exclusion where multiple candidates are excluded
together. It is commonly believed (and probably intended) that this can never alter the election
outcome. In this section we show that that assumption is mistaken.

The AEC does not appear to have implemented bulk exclusions since 2013, presumably in the
mistaken belief that it cannot not alter the count. It is unclear whether bulk exclusion is at the
AEC’s discretion or not. The Electoral Act Section 273 (13) states:

5The oﬃcial Distribution of Preferences is at https://www.aec.gov.au/Elections/federal_elections/2016/

files/wa-senate-dist-prefs-2016fe-special-count-07032017.pdf

3


--- Page 4 ---

(13) Where, after the counting of ﬁrst preference votes or the transfer of surplus votes
(if any) of elected candidates, no candidate has, or fewer than the number of candidates
required to be elected have, received a number of votes equal to the quota:

(a) the candidate who stands lowest in the poll must be excluded; or

(b) if a bulk exclusion of candidates may be eﬀected under subsection (13A), those

candidates must be excluded;

The logic seems unclear here—we are not lawyers and cannot say with conﬁdence whether “those
candidates must be excluded” would apply whenever (13A) permits it, but that would be one
plausible interpretation.

In this section we provide two simple examples to show that, in two diﬀerent ways, bulk
exclusion can alter the outcome. The ﬁrst concerns the order of eliminations, the second is due to
rounding. In each case, diﬀerent people are elected depending on whether the bulk exclusion rule
is applied.

3.1 Bulk Exclusions and order of elimination

Subsection 13A (d) (i) does not have a clause requiring candidate C to satisfy a similar condition
to 13A (b) (ii). This means that bulk exclusions can change the order of elimination. This can
readily change the outcome of an election. A simple example is shown here with 6 vacancies, and
votes as described in the table below:

Preference List Occurrences
L1, L2, W1, W3
L2
L3, L2, W1, W3
L4, W1, W2
L5, W4
L5, W5
L5, W6
L5, W7
L5
W7
W6
W5
W4
W3
W2
W2, W4
W2, W5
W2, W6
W1

2
3
4
6
1
2
3
4
10
400
400
400
400
400
388
6
4
2
400

Using the rules as applied by the AEC in 2013, as best we can reverse engineer, would result

in the distribution of preferences shown in table 2, with bulk exclusion occurring on count 1:

However, using the rules that seem to have been applied by the AEC in 2016, ignoring bulk ex-
clusions, the order of elimination changes, and candidate W3 is elected instead of W2. Furthermore
the order of election of candidates W4, W5, W6 and W7 is reversed, as shown in table 3

Curiously enough, using the rules apparently applied by the AEC in 2019, the same candidates

are elected, but in a diﬀerent order; see section 4.2.

3.2 Bulk Exclusions and rounding

It is more obvious that Bulk Exclusions can change rounding, and this can change the result of
the election. Rounding may seem like a small eﬀect, but in 2016 Tasmania, there were 285 votes
lost due to rounding, and the margin between the last two candidates was 143 votes.

Here is a very simple demonstration of how the presence or absence of bulk exclusion can change

rounding and thus who is elected in a 2 vacancy election.

4


--- Page 5 ---

Count W1 W2 W3 W4 W5 W6 W7
400
400

400

400

400

400

Exclude L1, L2

400

400

400

400

400

400

1
400
2 +2
402
3 +4
406

Exclude L3
4
Exclude L4
5
Exclude L5
6
Exclude W3

400
+6
406

400

406 (cid:8)(cid:8)(cid:72)(cid:72)400
−400
0

406

406

406

406

400

400

400

400

400

400

400

400
400
+1 +2 +3 +4
404
401

403

402

401

402

403

404

L1
(cid:1)(cid:65)2

L2
(cid:1)(cid:65)3
−2 −3
0

0

L3
4

(cid:1)(cid:65)4
−4
0

L5
20

20

20

L4
6

6

(cid:1)(cid:65)6
−6

0 (cid:26)(cid:26)(cid:90)(cid:90)20
−20
0

Exh. TV

+3
3

3

3
+10
13
+400
413

1

1

1

1

1

Table 2: Elected : W1, W2, W7, W6, W5, W4. Rules used : AEC2013. Elected candidates are
colored purple once they have a quota.

Count W1 W2 W3
400

400

400

W4 W5 W6 W7
400
400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

L3
4

L1
(cid:1)(cid:65)2

L2
3
−2 +2
5

0

(cid:1)(cid:65)4
+4 −4
0

9

1
2
Exclude L1
3
Exclude L3

Exclude L4
5
Exclude L2
6
Exclude L5
7
Exclude W2
8
Surplus W4

400
4 +6
406

400

406

400

406 (cid:8)(cid:8)(cid:72)(cid:72)400
−400
0

406

406

(cid:1)(cid:65)9
−9
0

400
+6
406

406

406

406

400

400

400

400

400

400

400
400
+1 +2 +3 +4
401
404
403
402
+6 +4 +2
405
406
407
−1407
406

405

404

406

404

L4
6

6

(cid:1)(cid:65)6
−6
0

L5
20

20

20

20

(cid:26)(cid:26)(cid:90)(cid:90)20
−20
0

Exh. TV

1

1

1

1

1

1

1/407

+3
3
+10
13
+388
401
+1407
402

Table 3: Elected : W1, W3, W4, W5, W6, W7. Rules used : AEC2016. Elected candidates are
colored purple once they have a quota. Superscripts are the number of papers corresponding to
the given number of votes.

5


--- Page 6 ---

Count
1
400
2 −100400
300

Q W1 W2
246
245

245

246

Surplus Q
3
Exclude L1, L2, L3
4
Exclude L1, L2, L3

300

300

245
+29
247

246

246

L3
3

L2
2

L1
1

Exh. Rounding TV

+03 +03 +03 +97391
97
+6
103

(cid:1)(cid:65)1
−1
0
−03 −03 −03

(cid:1)(cid:65)3
−3
0

(cid:1)(cid:65)2
−2
0

103

+3
3

3
−2
1

1/4

1

1/4

Table 4: Elected : Q, W1. Rules used : AEC2013. Elected candidates are colored purple once
they have a quota. Superscripts are the number of papers corresponding to the given number of
votes.

Count
1
400
2 −100400
300

Q W1 W2
246
245

245

246

L3
3

L2
2

L1
1

Exh. Rounding TV

3

+03 +03 +03 +97391
97
+1
98

2

3

2

(cid:1)(cid:65)1
−1
0
−03

3

3

(cid:1)(cid:65)2
−2
0
−03

(cid:1)(cid:65)3
−3
0
−03

98
+2
100

100
+3
103

103

+3
3

3

3

3

3

3

3

1/4

1

1/4

1

1/4

1

1/4

300

300

300

300

300

300

245
+03
245

245
+03
245

245
+03
245

246

246

246

246

246

246

Surplus Q
3
Exclude L1
4
Exclude L1
5
Exclude L2
6
Exclude L2
7
Exclude L3
8
Exclude L3

Table 5: Elected : Q, W2. Rules used : AEC2016. Elected candidates are colored purple once
they have a quota. Superscripts are the number of papers corresponding to the given number of
votes.

Preference List Occurrences
L3
L2
L1
W1
W2
Q, L3, W1
Q, L2, W1
Q, L1, W1
Q

3
2
1
245
246
3
3
3
391

Using bulk exclusion as in 2013 produces winners Q and W1 as shown in table 4. Not using

bulk exclusion as in 2016 produces winners Q and W2 as shown in table 5.

3.3 Bulk Exclusion summary and conclusion

These examples show that, as currently legislated, bulk exclusion can alter election outcomes,
though it has not done so in practice so far.

Bulk exclusion is an unnecessary complication for an electronic count. The ambiguity over
whether or not to use it could result in a contested election outcome that is diﬃcult or impossible
to resolve. The best way to avoid this is to remove bulk exclusion from the Act.

Recommendation 4. Remove Bulk Exclusion from the Act. This includes subsections (13A),
(13B), (13C), portions of (13AA), and several deﬁnitions in (29).

6


--- Page 7 ---

4 Ambiguity in the Act: termination of the count

There are three diﬀerent ways a candidate can be elected:

• by attaining a quota,

• by being the higher of the two continuing candidates when there is exactly one unﬁlled seat

(Subsection (17)).

• by being a continuing candidate when the number of unﬁlled seats equals the number of

continuing candidates, (subsection (18))

The Act does not explicitly deﬁne when the distribution of preferences ends—in particular, it
is unclear when subsections (17) or (18) apply. Subsection (18) aﬀects only order, not who gets
elected. However, subsection (17) can aﬀect who wins.

Commonwealth Electoral Act 1918, Section 273, (17)

In respect of the last vacancy for which two continuing candidates remain, the con-
tinuing candidate who has the larger number of votes shall be elected notwithstanding
that that number is below the quota, and if those candidates have an equal number of
votes the Australian Electoral Oﬃcer for the State shall have a casting vote but shall
not otherwise vote at the election.

Unlike a computer program, which (usually) has an implicit but well-deﬁned sequence of exe-
cution steps, the Electoral Act simply lists a number of things that could happen (in subsections
(9) through (18)) but does not specify which should be chosen when the conditions for two or more
have been met.

As a speciﬁc example, suppose one candidate (A) attains a quota (and hence a seat), leaving
two other candidates, B and C, competing for a single last seat. Should subsection (17) apply
immediately, electing whichever of B and C has a larger total, or should A’s surplus be distributed
ﬁrst? The legislation is not clear.

Subsection (14) states

(14) Any continuing candidate who has received a number of votes equal to or
greater than the quota on the completion of a transfer under subsection (13) or (15)
of ballot papers of an excluded candidate or candidates, as the case may be, shall be
elected, and, unless all the vacancies have been ﬁlled, the surplus votes (if any) of the
candidate so elected shall be transferred...

The question is whether, “all the vacancies have been ﬁlled,” because (17) applies immediately,
or instead (17) should not be applied until after this transfer has been completed. A similar
ambiguity arises over exclusions (subsection (15)): when a candidate is excluded, should subsection
(17) apply immediately, or should the excluded candidate’s preferences be distributed ﬁrst?

These decisions could aﬀect who wins, because preference distribution could alter the relative

rankings of the two continuing candidates.

4.1 Practical examples of when subsection (17) can alter the election

outcome

In practice, for all the examples we could ﬁnd, the AEC seems to apply (17) only after all transfers
have been completed. This seems like the right thing to do, and we believe the legislation should
be amended to reﬂect (what seems to be) the AEC’s practice. Indeed, subsection (17) could simply
be removed—it is unnecessary because the exclusion of the lower candidate immediately causes
subsection (18) to apply (one candidate and one seat remaining).

Resolving the ambiguity is important because this case arises frequently in practice. For exam-
ple, consider the special count of the 2016 South Australian Senate votes, conducted in 2017 after
the removal of Bob Day.6 The last steps are summarised in Table 6. At the end of Count 453, there
are three remaining candidates: McEwen (ALP), Gichuhi (FFP), and Burgess (ON). Burgess has
the least votes and is excluded, leaving only two candidates. If subsection (17) were applied imme-
diately, McEwen would win (with the higher tally). But after all the preferences are distributed,
Gichuhi wins. Similar examples arose in 2013 in NSW, Vic, WA and Tasmania.

6https://www.aec.gov.au/Elections/Federal_Elections/2016/files/sa-senate-dist-prefs-2016fe-special-count-13042017.

pdf

7


--- Page 8 ---

Count McEWEN GICHUHI BURGESS

453
Surplus KAKOSCHKE-MOORE
454–465
Exclude BURGESS

54099

52482

(cid:24)(cid:24)(cid:24)(cid:88)(cid:88)(cid:88)46684

65841

69442

0

Table 6: The last steps of the 2016 South Australian Senate special election count, conducted in
2017. If paragraph (17) was applied at count 453, as soon as Burgess was excluded, McEwen would
win instead of Gichuhi.

Count W1 W2 W3 W4 W5 W6 W7
400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

400

L3
4

L1
(cid:1)(cid:65)2

L2
3
−2 +2
5

0

(cid:1)(cid:65)4
+4 −4
0

9

400
+6
406

406

400

400

400

400

400

400

400
400
+1 +2 +3 +4
404
401

403

402

(cid:1)(cid:65)9
−9
0

406

400

406

401

402

403

404

1
2
Exclude L1
3
Exclude L3

Exclude L4
5
Exclude L2
6
Exclude L5
7
Exclude W2

400
4 +6
406

400

406

400

406 (cid:8)(cid:8)(cid:72)(cid:72)400

L4
6

6

(cid:1)(cid:65)6
−6
0

L5 Exh. TV
20

1

1

1

1

1

20

20

20

+3
(cid:26)(cid:26)(cid:90)(cid:90)20
3
−20 +10
13

0

13

Table 7: Elected : W1, W3, W7, W6, W5, W4. Rules used : AEC2019. Elected candidates are
colored purple once they have a quota.

Recommendation 5. Remove subsection (17), or clarify that it applies only after all papers from
surplus distributions and exclusions are transferred (which seems to be the AEC’s current practice).

4.2 Subsection 18

Subsection (18) does not aﬀect who wins, but may aﬀect their order—we do not know whether
this matters in any politically-relevant way.

In 2016, the AEC seemed to apply section 18 after ﬁnishing started exclusions and any surplus
distributions that may thereby arise. For example, in the Queensland 2016 election, at count 830,
candidate R. McGarvie was excluded, leaving 2 candidates and 2 seats. The exclusion was carried
out in full (11 counts), and C Ketter was discovered to have a quota, leaving 1 candidate (M
Roberts) and 1 vacancy. This candidate was not elected until count 841, when C Ketter’s surplus
was distributed. Similar events happened in Victoria (count 814), and NSW (count 1054). A
diﬀerent thing occurred in WA 2016 (with R. Cullerton excluded) where on count 535 K. Muir was
excluded, leaving 2 candidates and 2 seats. The ﬁrst step of the exclusion was performed, at the
end of which the remaining 2 candidates were both declared elected. Regardless, the transfer of
votes for the ﬁrst step of exclusion was always performed.

In 2019 the rules seemed to change and the AEC applied subsection (18) immediately after
identiﬁcation of a candidate to be excluded, before transferring any votes. In 2019 NSW, count
429, K. McCulloch is excluded. This leaves 2 candidates, 2 vacancies. The preference distribution
was aborted and no ballots were transferred in this count.

Recommendation 6. Clarify when subsection (18) applies.

The example used in Section 3.1 can be applied using the rules the AEC appeared to have used
in 2019, aborting an exclusion after identiﬁcation of the candidate to be excluded. This produces
yet another diﬀerent outcome from the same votes shown in table 7. This time only the order of
election changes, with the order of election of candidates W4, W5, W6 and W7 being reversed.

8


--- Page 9 ---

5 Conclusion

Everyone makes mistakes, and all software has bugs and security vulnerabilities.7 For election
software, the crucial thing is not to make the software perfect (which is impossible) but to ensure
an open and transparent process that allows the software, and its results, to be scrutinised and
double-checked. It is a good thing when errors and vulnerabilities are found and ﬁxed.

We have found one error (in the implementation of tiebreaking) in the AEC’s oﬃcial Senate
counting software. This needs to be corrected before the next election, and we would be happy to
help patch it if the source code is made openly available. Alternatively, the Electoral Act could be
amended to match the AEC’s current algorithm—it does not really matter how ties are broken,
as long as the algorithm is speciﬁed in advance. Modifying the Electoral Act to match the current
algorithm would also be aided by opening the source code.

The same applies to the other two ambiguities we identiﬁed in the Electoral Act: bulk exclusion
and the order of application of subsection (17). The legislation is unclear, and needs to be clariﬁed,
either to match the AEC’s current code (if it is made available) or independently. A system that
allows discretion over ambiguous rules, or runs a count diﬀerent from the speciﬁed rules, may result
in disputes at election time that are diﬃcult or impossible to resolve.

The purpose of this report is to avoid this outcome, by providing speciﬁc directions on correcting
the AEC’s code (Section 2), realigning the AEC’s code with the unclear legislation (Section 3),
and resolving ambiguitites in the legislation (Section 4). This should give very high conﬁdence
that the counting software is accurate and produces the unique outcome deﬁned by legislation.

It is easy to double-check the Senate counting software, because the software’s inputs (votes)
and outputs (distributions of preferences) are both publicly available. This allows for the open
public double-checking that is critical to democracy. (Of course it does not guarantee that all such
problems will be found before they aﬀect a count.) It is very helpful that the AEC makes this
information publicly available.

By contrast, it is not possible to double-check—so as to ﬁnd and ﬁx errors in—the software that
converts paper ballots to digitized preferences. This leaves the AEC without a way of providing
evidence of an accurate election outcome. A statistical audit of randomly-chosen paper ballots,
in the presence of scrutineers, would allow the digitized preferences to be double-checked against
their paper originals. This would give onlookers a chance to ﬁnd and ﬁx problems, or to see that
there was good evidence that the number of discrepancies was small.

6 Summary of Recommendations

Recommendation 1. When the preference data ﬁles for Senate votes are published, there should
be a rigorous statistical audit to check that they accurately reﬂect the paper ballots. This should be
conducted in a way that allows Scrutineers to check both the algorithms and the data.

Recommendation 2. Either:

• Correct the AEC’s oﬃcial code to follow the legislation for tie resolution, or

• Ask the AEC to publish the oﬃcial code, including its tiebreaking rules, and amend the

legislation to match it.

Recommendation 3. Formalize as legislation the apparent policy to, in the case of otherwise
unresolved ties, favour candidates lower on the ballot.

Recommendation 4. Remove Bulk Exclusion from the Act. This includes subsections (13A),
(13B), (13C), portions of (13AA), and several deﬁnitions in (29).

Recommendation 5. Remove subsection (17), or clarify that it applies only after all papers from
surplus distributions and exclusions are transferred (which seems to be the AEC’s current practice).

Recommendation 6. Clarify when subsection (18) applies.

7Even our software. We found the AEC tiebreaking error when we were re-implementing our version of the

Senate count and noticed we had made the same mistake.

9


--- Page 10 ---

References

[BCT+19] Michelle Blom, Chris Culnane, Vanessa Teague, Damjan Vukcevic, Andrew Conway,
Rajeev Gor´e, and Peter Stuckey. Submission to the inquiry into the conduct of
the 2019 federal election, 2019. https://www.aph.gov.au/DocumentStore.ashx?id=
a47d4582-b11f-4731-b87a-852ddab7f1c3&subId=670312.

[CBNT17] Andrew Conway, Michelle Blom, Lee Naish, and Vanessa Teague. An analysis of New
South Wales electronic vote counting. In Proceedings of the Australasian Computer Sci-
ence Week Multiconference, pages 1–5, 2017. https://arxiv.org/pdf/1611.02015.
pdf.

[CT]

[GL16]

Andrew Conway and Vanessa Teague.
Errors in the act’s electronic count-
ing code. https://github.com/SiliconEconometrics/PublicService/raw/master/
CountVotes/2020%20Errors%20In%20ACT%20Counting.pdf.

Rajeev Gor´e and Ekaterina Lebedeva. Simulating stv hand-counting by computers
In International Joint Conference on Electronic Voting,
considered harmful: Act.
pages 144–163. Springer, 2016.
https://openresearch-repository.anu.edu.au/
bitstream/1885/113153/2/01%20Gore%20and%20Lebedeva%20Simulating%20STV%
20hand-counting%202017.pdf.

[Wen10] Roland Wen. Online Elections in Terra Australis. PhD thesis, University of New South
Wales, Sydney, Australia, 2010. http://unsworks.unsw.edu.au/fapi/datastream/
unsworks:10365/SOURCE02.

10

This document, titled "Corrections to the AEC’s counting code, and recommended amendments to the Electoral Act, for Senate Counting & Scrutiny" by Dr Andrew Conway and A/Prof Vanessa Teague (August 31, 2021), identifies discrepancies between the Australian Electoral Commission (AEC)'s Senate election counting software and the requirements of the Electoral Act, along with ambiguities within the Act itself.

Key Issues Identified:
* Tie-breaking Inconsistency: The AEC's software applied an incorrect tie-breaking rule for 3-way ties (e.g., in the 2016 WA Senate election), leading to candidates being excluded in the wrong order, contrary to Electoral Act Section 273 31(b). This deviation could potentially alter election outcomes.
* Bulk Exclusion Omission: The AEC has not implemented bulk exclusions (Electoral Act Section 273 (13A)) since 2013, under the presumed, but mistaken, belief that it does not affect the count. The report demonstrates that bulk exclusions can, in fact, change election outcomes due to altered elimination order and rounding effects. The Act's language on bulk exclusion is ambiguous regarding its mandatory application.
* Ambiguity in Count Termination: The Electoral Act lacks clarity on when the count terminates, particularly concerning the simultaneous application of subsections 273 (17) (electing the higher of two remaining candidates for the last seat) and (18) (electing remaining candidates when seats equal continuing candidates). This ambiguity can affect who wins the last seat or the order in which candidates are declared elected.

Recommendations:
The report proposes several legislative and procedural changes:
* Correct the AEC's official code to comply with legislated tie-breaking rules, or alternatively, publish the code and amend the legislation to match it.
* Formalize the policy of favoring candidates lower on the ballot for otherwise unresolved ties.
* Remove bulk exclusion provisions from the Electoral Act to eliminate complexity and ambiguity.
* Clarify or remove subsection 273 (17) to ensure it applies consistently (preferably after all transfers are completed, aligning with AEC's current practice).
* Clarify when subsection 273 (18) applies.
* Implement a rigorous statistical audit of digitized preference data files against paper ballots to ensure accuracy and transparency.

Relevance to FOI Request:
This document is the "subsequent 2021 report by Conway and Teague" directly referenced in the FOI request overview. It provides the detailed findings concerning "accuracy, methodology, and identified issues with the Australian Electoral Commission's (AEC) Senate election counting algorithm and associated scrutiny processes." Specifically, it substantiates the FOI request's points regarding:
* The AEC's software using tie-breaking rules inconsistent with the Electoral Act.
* The non-implementation of bulk exclusions and their potential to alter outcomes.
* Ambiguities in the Electoral Act concerning the count's termination.
* It details the proposed "legislative and procedural recommendations," including publishing the AEC's code, correcting tie-breaking, and clarifying statutory definitions, which the AEC subsequently responded to. As such, it is central to understanding the discrepancies investigated by the FOI request.

AI Generated

Document 6 - Corrections to the Senate Counting algorithm [SEC=OFFICIAL] (A1861209).pdf (pdf)

Download file

--- Page 1 ---

From:
To:
Cc:
Subject:
Date:
Attachments:

Kristen Wild
Vanessa Teague
andrewstvcount@greatcactus.org; Tom Rogers; Thomas Ryan
RE: Corrections to the Senate Counting algorithm [SEC=OFFICIAL]
Tuesday, 19 October 2021 11:28:00 AM
18 10 2021 - Dr Andrew Conway and Prof. Vanessa teague - Senate Counting and Scrutiny.pdf
image001.gif
image002.gif
image003.jpg

Good morning

Please find the attached response from Mr Rogers.

Regards
Kristen

Kristen Wild | Personal Assistant to the Commissioner
Executive Leadership Team
Australian Electoral Commission

From: Vanessa Teague <Vanessa.Teague@anu.edu.au> 
Sent: Tuesday, 31 August 2021 3:40 PM
To: Tom Rogers 
Cc: andrewstvcount@greatcactus.org
Subject: Corrections to the Senate Counting algorithm

; Thomas Ryan 

CAUTION: This email originated from outside of the Australian Federal Government. Do not click links
or open attachments unless you recognise the sender and know the content is safe.

Dear Mr Rogers and Mr Ryan,

Please find attached our report on the AEC's Senate count.

We have found one instance in which the software breaks ties in a way that is not what is
specified in the Electoral Act. We have also noted some other ambiguities that need to be
resolved either by amending the Act or by otherwise committing to the exact counting
algorithm.

We would be very happy to help you correct the code - you can see our implementation here
https://github.com/AndrewConway/ConcreteSTV.  We could easily make the necessary
adjustments to the official code if it was similarly published online.

Also, we support recent proposals to audit randomly selected paper ballots at the end of the
digitizing process to give everyone evidence of the accuracy of the process. We would be
happy to help the AEC implement these measures securely and transparently.

Yours Sincerely,

Vanessa Teague and Andrew Conway.

Vanessa Teague

--- Page 2 ---

CEO, Thinking Cybersecurity Pty. Ltd.

A/Prof (Adj.), Research School of Computer Science
The Australian National University | Canberra ACT 2601
E: vanessa.teague@anu.edu.au

W: https://researchers.anu.edu.au/researchers/teague-v

CRICOS Provider #00120C

This document is an email chain initiated by Vanessa Teague and Andrew Conway (August 31, 2021) to Australian Electoral Commission (AEC) Commissioner Tom Rogers, formally submitting their report on the AEC's Senate count. In this submission, they highlight a finding that the AEC's software "breaks ties in a way that is not what is specified in the Electoral Act" and notes other ambiguities requiring resolution through legislative amendment or a clarified algorithm. They offer assistance with code correction and support auditing paper ballots. The document's later email (October 19, 2021) from Kristen Wild (AEC) to Vanessa Teague then forwards Mr Rogers' (AEC) official response to their report.

Relevance to FOI Request:
This document is directly relevant as it provides primary evidence of the 2021 Conway and Teague report's official submission to the AEC. It explicitly details one of the "significant discrepancies" identified by the report, namely the inconsistent tie-breaking rules, a core focus of the FOI request. Furthermore, it confirms the AEC's subsequent engagement and response to the report's findings, directly addressing the FOI's scope regarding identified issues with the counting algorithm and the AEC's actions in response.

AI Generated

AEC FOI Disclosure Log Archive

FOI Request LEX3029

Summary of FOI Request Documents

Documents from the FOI Request

Main Content from the FOI Request Documents

Summary of Key Findings and Progressive Critique:

FOI Request LEX3029, Schedule of Released Documents [PDF 79KB] (pdf)

LEX3029 documents [ZIP 1.82MB] (zip)

ZIP Contents

Document 1 - For Information_ Independent test of senate counts [SEC=UNCLASSIFIED] (A1861023)_Redacted.pdf (pdf)

Document 2 - Bowland's recounts for various scenarios... [SEC=UNCLASSIFIED] (A1861038)_Redacted.pdf (pdf)

Document 3 - Corrections to the Senate Counting algorithm [SEC=OFFICIAL] (A1861055)_Redacted.pdf (pdf)

Document 4 - Recommended Amendments Senate Counting And Scrutiny (A1860962).pdf (pdf)

Document 6 - Corrections to the Senate Counting algorithm [SEC=OFFICIAL] (A1861209).pdf (pdf)

Document 7 - 18 10 2021 - Dr Andrew Conway and Prof. Vanessa teague - Senate Counting and Scrutiny (A1860875).pdf (pdf)