AustLII Home | Databases | WorldLII | Search | Feedback

University of New South Wales Law Journal Student Series

You are here:  AustLII >> Databases >> University of New South Wales Law Journal Student Series >> 2022 >> [2022] UNSWLawJlStuS 32

Database Search | Name Search | Recent Articles | Noteup | LawCite | Author Info | Download | Help

Hodgson, Tom --- "Can You Teach An Old Doctrine New Tricks? An Analysis Of The Doctrine Of Unilateral Mistake In Relation To Contractual Errors Made By Computers" [2022] UNSWLawJlStuS 32; (2022) UNSWLJ Student Series No 22-32


CAN YOU TEACH AN OLD DOCTRINE NEW TRICKS? AN ANALYSIS OF THE DOCTRINE OF UNILATERAL MISTAKE IN RELATION TO CONTRACTUAL ERRORS MADE BY COMPUTERS
TOM HODGSON
INTRODUCTION

Technological advancements have enabled software programs and algorithms (‘computers’) to generate and execute contracts without any human involvement. However, the risk of computer made errors questions whether the general law can be incrementally developed in order to protect operators from mistakes with catastrophic consequences. Recently, unilateral mistake was unsuccessfully relied upon before the highest Singaporean Court in relation to transactions executed by computers at approximately 250 times market value.[1]

By critically analysing Quoine, this essay considers whether the Australian doctrine of unilateral mistake can provide protection in relation to computer created contracts. This essay is divided into four parts. First, Quoine is briefly outlined. Second, contract formation principles in relation to machine automation and the justifications for generally holding operators liable are discussed. Third, possible incremental changes to the doctrine are considered by evaluating the judgments in Quoine. Finally, legislative reform and ways contractual drafting could attempt to protect operators from such mistakes are explored.

Ultimately, this essay argues that unilateral mistake cannot be developed to protect operators, as such changes would unprincipledly shift the focus of the doctrine from procedural fairness to substantive fairness. Nonetheless, due to the possibility of extreme substantive unfairness, legislative reform is needed to protect operators.

The focus of this essay is on deterministic algorithms, which means that ‘pre-programmed rules’ are applied.[2] Non-deterministic algorithms, that have ‘a mind of their own’, are beyond the scope of this essay.[3]

BACKGROUND

Computer created contracts are becoming a significant part of commercial dealings in a variety of industries.[4] Such automation is particularly prevalent in the financial and insurance sectors, where considerable amounts of data must be processed.[5] Due to the complexity of these computers, errors are not uncommon, and they can have drastic consequences for their operators.[6] Therefore, considering whether operators should always be bound by such contracts raises important questions about the appropriate allocation of risk. In particular, whether the general law and statute should protect operators from certain risks.

In Singapore, the doctrine of unilateral mistake has been applied to transactions that took place online in which a mistake was made by a human and the non-mistaken party was also a human.[7] Whilst Chwee Kin Keong involved an employee error that created the absurdly low price on the online retail website,[8] this case can likely be applied to computers transacting with humans. If a computer made an obvious error, a human counterpart purchasing the product would still be aware that the price was absurdly low and could not reflect the intention of the other party. Therefore, the human would have constructive notice of the mistake, like in Chwee Kin Keong.[9]

However, the application of the doctrine to contracts formed without any human involvement raises new issues that were recently explored in Quoine.[10] Both Quoine, the operator of a cryptocurrency exchange platform, and B2C2, a trader, used algorithmic trading technology that required no human involvement.[11] Due to Quoine’s oversight, several necessary changes to its operating systems were not made, preventing the platform from accessing external cryptocurrency market data.[12] Consequently, a ‘chain of events’ occurred because Quoine was the primary market maker.[13] This meant that Quoine usually made continuous buy and sell orders that largely set prices on the platform. However, the inability to access market data prevented it from doing so and ultimately led to the extreme prices. Ultimately, B2C2 executed transactions at approximately 250 times market value.[14] The next day, when the parties became aware of the trades, Quoine cancelled the trades on the basis of a purported unilateral mistake.[15]

Despite automated contracting having been recognised for a considerable time,[16] the novel issue was raised as to how the doctrine of unilateral mistake should apply to contracts formed without human involvement.[17] In short, the majority applied ordinary principles governing contract formation and unilateral mistake, resulting in judgment in favour of B2C2.[18] Applying the Australian doctrine of unilateral mistake, this essay will critically evaluate this decision and consider if any arguments supporting an expansion of the doctrine to cover this novel situation have merit.

CONTRACT FORMATION

Before discussing the role of equity to rescind otherwise binding contracts, an anterior issue is to consider when contracts involving computers are formed. Since Thornton, legally binding offers have been cable of being made by automated machines on their operator’s behalf.[19] The principle from Thornton has been applied to software programs, where it was irrelevant that the operator was not aware of the exact offers that the program would make.[20] Similarly, contracts can be formed in Australia by computer programs or ‘other automated means’ without human review or intervention.[21] Therefore, operators will generally be bound when they convey computers as being able to contract on their behalf.[22]

Operators generally being bound by contracts could be justified on the basis that this incentivises them to minimise the likelihood of errors by engaging in diligent programming and monitoring.[23] Whilst operators may be incentivised in this way, this justification is not entirely satisfactory. Coding errors still frequently occur in commercial software, even when operators act diligently.[24]

Whilst there is a genuine risk of computer errors, holding operators bound ensures legal certainty by enabling counterparties to appreciate that legally binding offers have been made.[25] As a party cannot argue that no contract was ever created when contracts formed are slightly unfavourable, predictability and certainty can be maintained in commercial dealings.[26] Further, this approach is consistent with the well-established objective theory of contracts, where the actual subjective intention of the parties is generally irrelevant.[27] By allowing other parties to contract with their computer, the outward intention of the operator is for the computer to enter contracts on their behalf.[28]

Therefore, operators can be deemed to accept the risk that a contract could be executed outside the parameters that have, or should have been, established.[29] Operators are ultimately making commercial decisions to take on risks associated with automation in order to receive commercial benefits, such as more efficient processes.[30] When no inequal bargaining power exists, courts will generally accept the parties’ liberty to allocate risks among themselves.[31]

Whilst contractual freedom and certainty remain important elements of commercial contracting, seemingly valid contracts can still be vitiated on the basis of recognised equitable principles.[32] Notwithstanding that operators are generally deemed to accept risks associated with computers, it must be considered whether this is a sufficient basis to make them bound by even the most disadvantageous transactions.[33] Consequently, there can still be arguments that equity, through the doctrine of unilateral mistake, should provide some protection to operators.

UNILATERAL MISTAKE

When unilateral mistake was last before the High Court, the majority established two limbs required to make enforcement of the contract unconscionable.[34] First, one party entered the contract under ‘a serious mistake’ in relation to the contents of a fundamental contractual term (‘mistake limb’).[35] Second, the non-mistaken party either contributed to the mistake or had actual or constructive knowledge of the other party’s mistake and did not correct it prior to contract formation (‘knowledge limb’).[36]

The approach in Taylor has been applied many times in lower courts.[37] Nonetheless, the High Court stated the test narrowly to resolve the appeal before it.[38] As Taylor is not an exhaustive statement on what establishes unilateral mistake,[39] the doctrine can possibly be adapted in order for computer mistakes to receive some protection.

A Knowledge Limb

Despite expansion of the doctrine being possible, cases so far have always required the non-mistaken party to have had actual or constructive knowledge of the mistake.[40] If this traditional approach can be applied to contracts formed without human involvement, it must be determined whose state of mind is relevant and when knowledge is to be considered.

Central to the majority’s adoption of this traditional approach in Quoine was the ability to assess knowledge prior to contract formation based on the Thornton principle.[41] As an intention to be bound can be imputed prior to contract formation,[42] the majority considered that knowledge could also be imputed at the time of programming or any time before contract formation.[43] Consequently, the knowledge limb was framed as whether the programmer had ‘actual or constructive knowledge’ that the relevant parameters of the program ‘would only ever be accepted by a party operating under a mistake’.[44] Therefore, the program must be designed to benefit from that particular mistake.[45]

1 Artificiality

Lord Denning’s approach in Thornton is understandable, as it was consistent with well-established contract formation principles and ensured contractual certainty by finding an objective intention to be bound. Nonetheless, the basis of this approach is highly artificial, being the assumption that a ticketing machine is essentially ‘a booking clerk in disguise’.[46] In making this assumption, of course Lord Denning did not contemplate the wider operation of this approach. In particular, that technological advancements could result in an infinite range of potential transactions being generated by computers, extending far beyond a simple ticketing machine that merely offered car park services at a set price. Therefore, in considering the novel issue in Quoine, the use of the Thornton principle as the basis to impute knowledge in relation to unilateral mistake can be questioned.

Notably, the majority’s approach essentially prevented any operation of unilateral mistake in this context. As programmers are unable to predict the future of markets and the endless potential coding errors,[47] they are unlikely to ever satisfy the knowledge requirement. Even the majority conceded that their approach is ‘artificial, even unrealistic’, as programmers are not ‘prophets’ and possible mistakes are wide ranging.[48]

Black swan events, that may be very rare but carry drastic consequences, will be outside the scope of the doctrine, as programmers will not contemplate the ‘real possibility’ of such events occurring.[49] The trial judge in Quoine concluded that B2C2’s programmer, despite inserting deep prices that ultimately allowed the transactions to occur, did not consider there to be a ‘real possibility’ of such transactions occurring.[50] The deep prices were inserted to safeguard B2C2 from ‘the unexpected happening’ and the programmer did not genuinely consider the possibility of Quoine making such an error.[51] Therefore, programmers are likely to attempt to plan for all market conditions, rather than seek to take advantage of others’ mistakes.[52]

Whilst B2C2’s software was only programmed by one person, who was also a company director, creating complex programs often requires the work of many programmers over an extended period of time.[53] Such programmers will often be merely following instructions from clients or managers.[54] Further, parties are increasingly reliant on ‘off the shelf’ algorithms that are purchased from external companies.[55] Therefore, there may also be very difficult questions of fact in determining whose knowledge should even be evaluated.[56]

I acknowledge the majority’s argument that their test enabled consideration of ‘the actual conduct of the parties’, as knowledge after programming up until contract formation can be considered.[57] For example, when the programmer subsequently became aware of a likely mistake but did not alter the program.[58] However, as counterparties in commercial transactions are unlikely to alert others to their own software issues, I question whether this would ever realistically happen. As the facts in Quoine highlight, extreme mistakes can occur that neither party genuinely considered could occur. Further, if third-party programmers are involved, they may be disconnected from the actual circumstances relevant to the transactions.[59]

Whilst parties can be deemed to accept risks associated with automation, should they also be prevented from ever relying on vitiating factors?[60] The majority’s traditional approach to unilateral mistake, as they found it possible to impute knowledge, reflected a failure to consider this issue. Therefore, imputing knowledge prior to contract formation, on the basis of the artificial principle in Thornton, is an entirely unsatisfactory approach.

B Mistake Limb

Similarly, the case law has not departed from the requirement in Taylor that there must be a serious mistake in relation to a fundamental contractual term.[61] However, serious mistakes in a computer context can potentially occur outside this traditional conception. For example, in Quoine there was no mistake in relation to a contractual term, as the computers operated as programmed.[62] The only possible mistake was the manner that Quoine’s platform acted in response to Quoine’s failure to update ‘critical operating systems’, which led to the ultimate transactions.[63] Nonetheless, the mistake and transactions subsequently executed could be argued to be as fundamental ‘in human terms’ as any other mistake protected by the doctrine.[64]

Whilst the transactions were ultimately due to Quoine’s omissions, fault does not necessarily prevent equitable intervention for unilateral mistakes.[65] The majority in Taylor referred to the US position, with seemingly approval, that the doctrine can still be available when the mistake was caused by ‘negligence or want of care’.[66] Further, fundamental errors similar to that which occurred in Quoine often occur ‘without any fault’.[67]

INCREMENTAL CHANGE?

Due to the impracticality of the majority’s approach and the potential for equitable principles to adapt to new circumstances, one could argue that unilateral mistake must be incrementally developed. As noted above, new circumstances can potentially establish unconscionability.[68] Whilst past cases only considered the knowledge limb at a point in time prior to contract formation,[69] there was no reason for a different approach. Knowledge could be derived from parties’ actual interactions leading up to contract formation. However, the conventional approach to knowledge cannot be applied when humans are not involved in contract formation.[70]

Therefore, the inapplicability of the conventional approach can be seen as a fundamentally new circumstance which requires a flexible response from equity, in which knowledge is considered after contract formation. Consequently, Mance IJ argued that transactions should be voidable if a reasonable person in the circumstances of the parties would have immediately thought the transactions were caused by ‘a fundamental mistake’.[71] Mance IJ held that the mistake would have been ‘obvious’ to any reasonable person, as B2C2 emailed Quoine at 6:15am the next day stating ‘Major Quoine database breakdown’ and asking to be called urgently.[72]

Mance IJ saw the underlying rationale of the doctrine as not being ‘a lack of correspondence between offer and acceptance’ but rather ‘a principle of justice’.[73] Therefore, Mance IJ focused on substantive unfairness, being that it is unconscionable for the non-mistaken party to take advantage of an extreme mistake upon learning about the mistake.[74] Due to the artificiality of the majority’s approach and the reasonable possibility of computer made mistakes, this novel situation could be considered a new form of ‘special circumstances’ that considers unconscionability after contract formation.[75] In extreme cases, where the mistake is so obvious that any reasonable trader would appreciate the mistake, it is unconscionable for the non-mistaken party to benefit from the contract upon learning about the fundamental error.[76]

A Procedural Fairness

Whilst no doubt a radical approach, one could argue that Mance IJ’s test is consistent with the fundamental principles underlying the doctrine. In particular, that it is unconscionable for the non-mistaken party to enforce the contract.[77] Further, as the common law fails to preserve commercially suitable outcomes in this context, it is consistent with the general role of equity to ensure just outcomes.[78]

However, notwithstanding the possibility of very substantive unfairness, vitiating factors seek to protect parties from entering into contracts that are procedurally unfair.[79] Unconscionability is not a free-standing doctrine and equity will not intervene on the basis of ‘some indefinable concept of unfairness’ outside of the recognised vitiating factors.[80] Due to this procedural focus, the fundamental concern of unilateral mistake is ‘whether it is unconscionable’ for the non-mistaken party to ‘knowingly take advantage’ of the other party’s mistake.[81] One party merely making a mistake, ‘no matter how serious’, is insufficient on its own to establish unconscionability.[82]

Even though B2C2 benefitted by millions of dollars from the errors, it would be unprincipled for equity to set aside the transactions because B2C2 had a clear conscience when the transactions were executed.[83] Of course, the non-mistaken party will essentially always have a clear conscience when computers are contracting on their behalf. However, this reason alone could not justify courts departing from the well-established principle that equity will not intervene when a party has a clear conscience.[84] Therefore, courts would not be able to depart from the requirement that the non-mistaken party must know about the other party’s mistake ‘before the transaction is completed’.[85]

Similarly, the distinct doctrine of unconscionable conduct seeks to preserve procedural fairness.[86] Unconscionability is not established merely because a transaction is unfair or unjust.[87] Whilst unconscionability is a broad concept, judges do not have a general power to set aside any contracts simply because they perceive the bargain to be ‘unfair, harsh or unconscionable’.[88] Therefore, Mance IJ’s focus on substantive fairness is not supported by recognised conceptions of unconscionability.[89]

1 Mistake Limb

Future mistakes within this context may relate to the fundamental terms of the contract, as required.[90] However, when they do not, like in Quoine, arguing that the mistake limb should be expanded is also likely unprincipled for similar reasons to the arguments above.

I acknowledge that in Singapore it has been left open in obiter whether unilateral mistake in equity could address ‘a wider and perhaps open-ended category of ‘fundamental’ mistake’.[91] Nonetheless, it is difficult to conceive that information external to transactions, such as the failure to update operating systems, could be considered. In no cases has there been a consideration of mistakes extraneous to the contract.[92] For example, the object of rectifying a contract on the basis of unilateral mistake is to correct a ‘disconformity’ between the terms of the contract and the intention of the parties.[93] In Quoine, the contractual terms reflected the intention of the parties, as the computers operated as programmed.[94] Therefore, expansion to cover such mistakes also illegitimately focuses on substantive fairness.

REFORM

Ultimately, Mance IJ’s position unprincipledly retracted from equity’s focus on procedural fairness.[95] Nonetheless, the inability of equity to protect parties from extreme substantive unfairness questions whether legislative reform is needed to preserve commercially fair outcomes. As coding errors inevitably occur in relation to commercial software,[96] operators should have some legal protection.

A Contractual Drafting

In Quoine, the parties chose to transact without human involvement, and they could have included a contractual clause that sought to mitigate the potential consequences of computer errors.[97] Therefore, one could argue that reform is not needed, as parties can adequately safeguards themselves from such risks if they want to. However, the complexity of these computers creates uncertainty as to whether contractual safeguards can sufficiently protect against possible errors.

For example, the terms of use on a trading platform could state:

If Technology used to generate and/or execute a transaction makes an Error in relation to a transaction, either party may terminate the transaction by written notice within 48 hours of the transaction being executed.

...

Error means when either or both parties’ Technology does not operate within the programmed parameters, resulting in a transaction being at least five times higher or lower than the Market Price.[98]

In Quoine, the software acted entirely as programmed and yet there was such a catastrophic outcome for one party.[99] Quoine unsuccessfully argued that a Risk Disclosure Statement, which included an aberrant value clause, was incorporated into the contract. This clause was found to be incompatible with a contractual clause that stated trades were irreversible.[100] Nonetheless, the majority still construed the possible effect of the aberrant value clause, highlighting the difficulties of drafting in this context. This clause stated:

The system may produce an aberrant value for the buy or sell price of the virtual currency calculated by the system. Please be aware that if the Company finds that a transaction took effect based on an aberrant value, the Company may cancel the transaction.

For example, the orders placed by Quoine were not ‘aberrant’, as they were generated exactly as programmed.[101] The alleged ‘malfunction’ was the platform’s inability to access external data due to Quoine’s failure to update operating systems and not due to software errors.[102]

Therefore, the above drafted clause would face similar issues. Of course, a more detailed contractual clause could set out all the circumstances in which a mistake is defined to occur. However, Quoine highlights that types of errors may not be predicted at the time of drafting.

An alternative solution could be to include a clause that adopted the wording of Mance IJ’s test, enabling a wide range of mistakes to be protected against. However, issues of interpretation could still arise. Whilst the outcome of the alleged error was so extreme in Quoine, where do you draw the line between acceptable mistakes and those considered to be obviously ‘fundamental’? Nonetheless, parties could clearly define ‘fundamental’ by, for example, a minimum percentage that transactions must differ from market value. The chosen percentage would depend on the risk that the parties are willing to take and whether such protection is only for very extreme outcomes.

In any event, the viability of contractual drafting as a solution can be further questioned as one party will likely have far greater negotiating power then the other. For example, Quoine controlled the platform and therefore could contract out certain risks that were in its interests.[103] Platform operators, such as Quoine, could likely offer contractual terms ‘on a take it or leave it basis’ to traders, regardless of whether the traders were sophisticated companies or individuals.[104] Therefore, if a trader’s error resulted in extremely disadvantageous transactions, they would be unlikely to have any possible contractual protection. Ultimately, contractual drafting alone is an insufficient solution.

B Legislation

Notwithstanding courts’ reluctance to consider substantive fairness, parliament has implemented legislation to address inequalities in contract law on multiple occasions.[105] For example, in standard form consumer and small business contracts, terms deemed to be substantially unfair are void in certain circumstances.[106] Further, statutory unconscionable conduct provisions enable consideration of both procedural and substantive unfairness, including the terms of the contract.[107] Therefore, legislation based on substantive unfairness is not completely novel.

I acknowledge that the provisions noted seek to protect particularly vulnerable parties. For example, parties required to accept standard form contracts on ‘a take it or leave it basis’.[108] In contrast, those using computers to contract will likely be sophisticated entities, at least while computer contracting technologies remain quite new. However, as commercial software errors will inevitably occur,[109] operators may be unable to protect themselves from extreme unfairness. The consequences of mistakes can be millions of dollars, like in Quoine,[110] and the possibility of a company going insolvent or other drastic consequences is not out of the picture. Therefore, parties in this context are often also vulnerable, supporting the need for legislative protection.

Further, legislative protections do not only safeguard unsophisticated individuals. There is also protection for businesses and the types of businesses that protections apply to appears to be growing. For example, there is currently a bill in relation to the standard form contract protections that would enable the protections to also be available for moderate sized businesses that could be considered sophisticated.[111] Under this bill, those who the provisions apply to will be increased from less than 20 employees to ‘fewer than 100’ and an alternative eligibility threshold, being when annual turnover is less than $10,000,000, would also be added.[112] Therefore, many technology companies that may use automated contracts, such as start-ups and even more developed businesses, are capable of receiving these protections, illustrating how creating new legislation to predominantly protect such businesses is not inconceivable.

Legislative protection could be achieved by amending the Electronic Transactions Act (‘the Act’).[113] Mance IJ’s test could be the basis of the new provision,[114] as operators should not be deemed to accept the risk of extremely unfair transactions that are obviously caused by fundamental mistakes.[115] Section 15D of the Act already provides protection for human made mistakes when contracting with computers.[116] This protection is needed, as the doctrine of unilateral mistake could not protect persons from situations covered by section 15D, as the non-mistaken party is a computer. The proposed reform can be seen as a logical and necessary subsequent amendment, as section 15D protects humans from mistakes made when contracting with computers in an area that the general law provides no protection. Contracts without any human involvement are becoming increasingly prevalent and therefore operators should be protected in this new computer context. In section 15D and the reform proposed, both provisions seek to ensure parties receive procedural fairness, preventing situations in which one party gains an extremely unfair advantage as a result of a mistake.

This amendment is also consistent with the purpose of the Act, being to maintain ‘business and community confidence’ in electronically executed contracts.[117] The reform would preserve this confidence by safeguarding operators from extreme unfairness and therefore encouraging the use of automated contracts. Further, commercial sensibility would be preserved, as the non-mistaken party does not deserve to receive such a benefit. As autonomous contracting is becoming a fundamental part of commercial dealings with major benefits, such as economic efficiencies and reductions in transaction costs,[118] maintaining this confidence is crucial for the modern economy. Ultimately, this reform would help encourage innovation and the use of new technologies, helping Australia to be a world leader in this space.

In order to encourage operators to act with reasonable levels of precaution and diligence, perhaps operators should not be protected for complete failures to safeguard themselves from errors.[119] Nonetheless, operators should only be expected to implement reasonable mechanisms and not to address every possible risk.[120] Even though Quoine caused the errors, there is nothing to suggest its conduct was ‘so egregious’.[121]

Additionally, the legislative protection should only be available in extreme circumstances. If protection was not limited to fundamental mistakes that were obvious, contractual certainty would be undermined as a constant risk would be that counterparties would seek to set aside slightly unfavourable transactions.

Of course, there would be contested cases where drawing the line between acceptable mistakes and those ‘fundamental’ is difficult. Nonetheless, this can somewhat be mitigated by having specific criteria for this enquiry. In particular, there could be minimum thresholds that transactions generally must differ from market value to make the error fundamental. These minimum thresholds could be set differently depending on the relevant industry, the type of transaction and the characteristics of the parties, such as their size. For example, perhaps for cryptocurrency or stock exchanges, the minimum threshold could be ten times the market value for the particular cryptocurrency or stock at the time of the mistake on the relevant platform. Other specific categories, such as insurance, could have a differently determined threshold. Alternatively, there could be general thresholds for all transactions, with a list of factors that courts could consider if a decision was litigated. I encourage further consideration of this at a future time, taking into account the perspectives of stakeholders that use automated contracts.

Further, greater contractual certainty could be achieved by including a time limit during which the party that made the mistake is able to contact the non-mistaken party to rescind the contract. For example, providing written notice to the counterpart within 48 hours of the mistake. I believe 48 hours is sufficient, as the mistaken party will likely identify extreme mistakes very quickly, as occurred in Quoine.[122]

Ultimately, the inability of the general law to preserve commercially fair outcomes in this context supports the need for legislation to protect parties against mistakes with catastrophic consequences.

CONCLUSION

Contracts entered into by computers without human involvement raise interesting questions about the appropriate allocation of risk, as mistakes will inevitably occur. Operators should generally be deemed as accepting risks associated with the commercial decision to automate, as this ensures contractual certainty and respects parties’ freedom to allocate risks. However, enforcing contracts that are obviously the result of errors that neither party ever expected could occur is another thing entirely.

Whilst equity provides flexibility to the law by making otherwise binding contracts voidable, equity is based on principle and not freestanding notions of fairness. As the current law on unilateral mistake cannot protect operators from fundamental errors, there is a genuine possibility that extreme substantive fairness will occur, like in Quoine.

Ultimately, due to the general law’s inability to sufficiently adapt to contracts without human involvement, legislation is needed to ensure operators are protected from substantially unfair results. Whilst contracts without human involvement remain quite novel, such contracts will only become more prevalent. The law must proactively grapple with such issues as they arise.

I encourage future research to consider in more depth legislative reform, including the perspectives of different stakeholders. I also encourage consideration of how non-deterministic algorithms, that have ‘a mind of their own’, should be legally regulated.


[1] Quoine Pte Ltd v B2C2 Ltd [2020] SGCA(I) 02 (‘Quoine’).

[2] Johnathan Tan Ming En, ‘Non-Deterministic AI Systems and the Future of the Law on Unilateral Mistakes in Singapore’ (2022) 34 SAcLJ 91, 93.

[3] Ibid.

[4] Ibid 92; Harry Surden, ‘Computable Contracts’ (2012) 46 UC Davis Law Review 629, 695.

[5] Johnathan Tan Ming En, ‘Non-Deterministic AI Systems and the Future of the Law on Unilateral Mistakes in Singapore’ (2022) 34 SAcLJ 91, 92.

[6] Eliza Mik, 'From Automation to Autonomy: Some Non-existent Problems in Contract Law' (2020) 36 Journal of Contract Law 205, 225.

[7] Chwee Kin Keong and others v Digilandmall.com Pte Ltd [2005] 1 SLR(R) 502.

[8] Chwee Kin Keong and others v Digilandmall.com Pte Ltd [2005] 1 SLR(R) 502.

[9] Chwee Kin Keong and others v Digilandmall.com Pte Ltd [2005] 1 SLR(R) 502.

[10] Quoine (n 1).

[11] Quoine (n 1) [1] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[12] Ibid [27]-[28] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[13] Ibid [2], [27]-[28] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[14] Ibid [2], [27] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[15] Ibid [2], [4] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[16] Thornton v Shoe Lane Parking [1970] EWCA Civ 2; [1971] 2 QB 163.

[17] Quoine (n 1) [78] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[18] Ibid [78]-[128] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[19] Thornton v Shoe Lane Parking Ltd [1970] EWCA Civ 2; [1971] 2 QB 163, 169.

[20] R (on the application of Software Solutions Partners Ltd) v Her Majesty’s Commissioners for Customs and Excise [2007] EWHC 971 (Admin) [65], [67] (‘Software Solutions’).

[21] Electronic Transactions Act 1999 (Cth) ss 5(1), 15C.

[22] Software Solutions (n 16) [65], [67].

[23] Eliza Mik, 'From Automation to Autonomy: Some Non-existent Problems in Contract Law' (2020) 36 Journal of Contract Law 205, 223, 226.

[24] Kelvin FK Low and Eliza Mik, ‘Lost in Transmission: Unilateral Mistakes in Automated Contracts’ (2020) 136 Law Quarterly Review 563, 568.

[25] Eliza Mik, 'From Automation to Autonomy: Some Non-existent Problems in Contract Law' (2020) 36 Journal of Contract Law 205, 223.

[26] Ibid 220.

[27] Commissioner of Taxation v Carter [2022] HCA 10; (2022) 399 ALR 521 [44] citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 429; Toll (FGCT) Pty Ltd v Alphapharm Pty Ltd (2004) 219 CLR 165 [41] citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 429.

[28] Software Solutions (n 16) [65]; Eliza Mik, 'From Automation to Autonomy: Some Non-existent Problems in Contract Law' (2020) 36 Journal of Contract Law 205, 217- 218.

[29] Software Solutions (n 16) [2], [65].

[30] Eliza Mik, 'From Automation to Autonomy: Some Non-existent Problems in Contract Law' (2020) 36 Journal of Contract Law 205, 217, 223, 226.

[31] Anthony Gray, ‘Unfair Contract Terms: Termination for Convenience’ [2013] UWALawRw 12; (2013) 37(1) University of Western Australia Law Review 229, 231-233.

[32] Toll (FGCT) Pty Ltd v Alphapharm Pty Ltd (2004) 219 CLR 165 [41] citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 429; Lexis Advance, Halsbury’s Laws of Australia (online at 26 February 2020) Vitiating Factors, ‘Limits on general law relief against harsh contracts’ [110-5880].

[33] Quoine (n 1) [193] (Mance IJ).

[34] Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 432 (Mason ACJ, Murphy and Deane JJ).

[35] Ibid.

[36] Ibid.

[37] For example, see Tutt v Doyle (1997) 42 NSWLR 10; Council of the City of Sydney v Wilson Parking Australia Pty Ltd [2015] NSWLEC 42 [64]-[67]; Westpork Pty Ltd v Bio-Organics Pty Ltd [2018] WASC 291 [116]; Schwartz Family Co Pty Ltd v Capitol Carpets Pty Ltd [2019] NSWSC 238 [67]-[68]; Reilly v Australia and New Zealand Banking Group Limited (No 2) [2020] FCA 1502 [138]-[139].

[38] Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 432 (Mason ACJ, Murphy and Deane JJ).

[39] Schwartz Family Co Pty Ltd v Capitol Carpets Pty Ltd [2019] NSWSC 238 [69] citing Tutt v Doyle (1997) 42 NSWLR 10, 14 and Blackley Investments Pty Ltd v Burnie City Council (No 2) [2011] TASFC 6; (2011) 21 Tas R 98 [10]; Reilly v Australia and New Zealand Banking Group Limited (No 2) [2020] FCA 1502 [142]-[143].

[40] For example, see Prudential Assurance Co Ltd v CM Breedon Pty Ltd [1994] VicRp 69; [1994] 2 VR 452, 457; Tutt v Doyle (1997) 42 NSWLR 10, 14; Mohamed v Farah [2004] NSWSC 482 [56]-[57]; Blackley Investments Pty Ltd v Burnie City Council (No 2) [2011] TASFC 6; (2011) 21 Tas R 98 [9]; Council of the City of Sydney v Wilson Parking Australia Pty Ltd [2015] NSWLEC 42 [66].

[41] Under Singaporean law, there is unilateral mistake at common law and in equity. Nonetheless, there is no need to distinguish the tests for the purposes of this essay. At common law, the non-mistaken party must have actual knowledge. In equity, knowledge only needs to be constructive, but the conduct must be unconscionable: Quoine (n 1) at [81].

[42] Thornton v Shoe Lane Parking Ltd [1970] EWCA Civ 2; [1971] 2 QB 163, 169; Software Solutions (n 16) [65], [67].

[43] Quoine (n 1) [104] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[44] Ibid [98], [103]-[104] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[45] Ibid [103]-[104] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[46] Thornton v Shoe Lane Parking Ltd [1970] EWCA Civ 2; [1971] 2 QB 163, 169.

[47] Kelvin Low and Eliza Mik, ‘Unpicking a Fin(e)tech Mess: Can Old Doctrines Cope in the 21st Century?’ Oxford Business Law Blog (Blog, 8 November 2019) <https://www.law.ox.ac.uk/business-law-blog/blog/2019/11/unpicking-finetech-mess-can-old-doctrines-cope-21st-century>.

[48] Quoine (n 1) [102] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[49] Allen Kiat Peng, ‘Contract Formation and Mistake in Cyberspace (Again)’ (2021) 33 SAcLJ 692,

720; Journe Fu, ‘Algorithmic Contracts: Who is to Blame?’ (2021) 12 Singapore Law Review: Juris Illuminae 1, 6.

[50] Quoine (n 1) [189] (Mance IJ) citing B2C2 Ltd v Quoine Pte Ltd [2019] SGHC(I) 03 [220]-[230].

[51] Ibid.

[52] Kelvin FK Low and Eliza Mik, ‘Lost in Transmission: Unilateral Mistakes in Automated Contracts’ (2020) 136 Law Quarterly Review 563, 567.

[53] Ibid.

[54] Ibid.

[55] Journe Fu, ‘Algorithmic Contracts: Who is to Blame?’ (2021) 12 Singapore Law Review: Juris Illuminae 1, 7.

[56] Ibid 6.

[57] Quoine (n 1) [99], [104] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[58] Ibid [99] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[59] Journe Fu, ‘Algorithmic Contracts: Who is to Blame?’ (2021) 12 Singapore Law Review: Juris Illuminae 1, 7.

[60] Kelvin Low and Eliza Mik, ‘Unpicking a Fin(e)tech Mess: Can Old Doctrines Cope in the 21st Century?’ Oxford Business Law Blog (Blog, 8 November 2019) <https://www.law.ox.ac.uk/business-law-blog/blog/2019/11/unpicking-finetech-mess-can-old-doctrines-cope-21st-century>.

[61] For example, see Schwartz Family Co Pty Ltd v Capitol Carpets Pty Ltd [2019] NSWSC 238 [74] citing Blackley Investments Pty Ltd v Burnie City Council (No 2) [2011] TASFC 6; (2011) 21 Tas R 98 [13].

[62] Quoine (n 1) [114] (Menon CJ, Boon Leong JA, Prakash JA and French IJ), [182] (Mance IJ).

[63] Ibid.

[64] Ibid [114], [182] (Mance IJ).

[65] Ibid [195] (Mance IJ).

[66] Ibid [195] (Mance IJ) citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 431; Seward, In the Marriage of (1984) 9 Fam LR 867, 871 citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 431; Commonwealth v VL Investments Pty Ltd [1988] ANZ ConvR 150 citing Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 431.

[67] Quoine (n 1) [196] (Mance IJ).

[68] Tutt v Doyle (1997) 42 NSWLR 10, 14.

[69] For example, see Prudential Assurance Co Ltd v CM Breedon Pty Ltd [1994] VicRp 69; [1994] 2 VR 452, 457; Tranchita v Retravision (WA) Pty Ltd [2001] WASCA 265 [42]; Mohamed v Farah [2004] NSWSC 482 [56]-[57]; Council of the City of Sydney v Wilson Parking Australia Pty Ltd [2015] NSWLEC 42 [66].

[70] Quoine (n 1) [185] (Mance IJ).

[71] Ibid [178], [204] (Mance IJ).

[72] Ibid [153] (Mance IJ).

[73] Ibid [181] (Mance IJ).

[74] Ibid [193], [205] (Mance IJ).

[75] Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 431 (Mason ACJ, Murphy and Deane JJ).

[76] Ibid [205] (Mance IJ).

[77] Ibid 431 (Mason ACJ, Murphy and Deane JJ) citing Torrance v Bolton [1872] UKLawRpCh 110; (1872) LR 8 Ch App 118.

[78] Quoine (n 1) [193], [205] (Mance IJ).

[79] Lance Rundle, ‘Thorne v Kennedy — A reminder by the High Court of Australia the law of contract and equity underpin family law financial agreements’ (2020) 48 Australian Bar Review 405, 416 citing Wilton v Farnworth [1948] HCA 20; (1948) 76 CLR 646, 655 (Rich J).

[80] Lexis Advance, Halsbury’s Laws of Australia (online at 26 February 2020) Vitiating Factors, ‘Limits on general law relief against harsh contracts’ [110-5880].

[81] Schwartz Family Co Pty Ltd v Capitol Carpets Pty Ltd [2019] NSWSC 238 [70] citing Tutt v Doyle (1997) 42 NSWLR 10, 12, 14.

[82] P Ward Civil Engineering Pty Ltd v Civil and Civic Pty Ltd [1999] NSWSC 727 [397].

[83] Patrick Stevedores Operations (No 2) Pty Limited v Port of Melbourne Corporation [2016] VSC 528 [63] citing Casquash Pty Ltd v NSW Squash Ltd (No 2) [2012] NSWSC 522 [7] and P Ward Civil Engineering Pty Ltd v Civil and Civic Pty Ltd [1999] NSWSC 727 [397].

[84] Ibid.

[85] Ibid.

[86] Lance Rundle, ‘Thorne v Kennedy — A reminder by the High Court of Australia the law of contract and equity underpin family law financial agreements’ (2020) 48 Australian Bar Review 405, 416 citing Wilton v Farnworth [1948] HCA 20; (1948) 76 CLR 646, 655 (Rich J).

[87] Sgargetta v National Australia Bank Ltd [2014] VSCA 159 [101] citing Attorney-General (NSW) v World Best Holdings Ltd (2005) 63 NSWLR 557 [120] (Spigelman CJ).

[88] Louth v Diprose (1992) 175 CLR 362, 405 (Toohey J).

[89] Vincent Ooi and Kian Peng Soh, ‘Rethinking Mistake in the Age of Algorithms: Quoine Pte Ltd v B2C2 Ltd’ (2020) 31(3) King's Law Journal 367, 370.

[90] Taylor v Johnson [1983] HCA 5; (1983) 151 CLR 422, 432 (Mason ACJ, Murphy and Deane JJ).

[91] Quoine (n 1) [91] (Menon CJ, Boon Leong JA, Prakash JA and French IJ) citing Chwee Kin Keong and others v Digilandmall.com Pte Ltd [2005] 1 SLR(R) 502 [75].

[92] For example, see Schwartz Family Co Pty Ltd v Capitol Carpets Pty Ltd [2019] NSWSC 238 [74] citing Blackley Investments Pty Ltd v Burnie City Council (No 2) [2011] TASFC 6; (2011) 21 Tas R 98 [13].

[93] Vantage Systems Pty Ltd v Priolo Corporation Pty Ltd [2015] WASCA 21 [164]-[165].

[94] Quoine (n 1) [114] (Menon CJ, Boon Leong JA, Prakash JA and French IJ), [182] (Mance IJ).

[95] Allen Kiat Peng, ‘Contract Formation and Mistake in Cyberspace (Again)’ (2021) 33 SAcLJ 692, 693; Vincent Ooi and Kian Peng Soh, ‘Rethinking Mistake in the Age of Algorithms: Quoine Pte Ltd v B2C2 Ltd’ (2020) 31(3) King's Law Journal 367, 370.

[96] Kelvin F.K. Low and Eliza Mik, ‘Lost in Transmission: Unilateral Mistakes in Automated Contracts’ (2020) 136 Law Quarterly Review 563, 568.

[97] Quoine (n 1) [104] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[98] Technology and Market Price would also be defined terms.

[99] Quoine (n 1) [69] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[100] Ibid [71] (Menon CJ, Boon Leong JA, Prakash JA and French IJ), [161] (Mance IJ).

[101] Ibid [69] (Menon CJ, Boon Leong JA, Prakash JA and French IJ).

[102] Ibid.

[103] Allen Kiat Peng, ‘Contract Formation and Mistake in Cyberspace (Again)’ (2021) 33 SAcLJ 692, 717.

[104] Ibid 718.

[105] Anthony Gray, ‘Unfair Contract Terms: Termination for Convenience’ [2013] UWALawRw 12; (2013) 37(1) University of Western Australia Law Review 229, 246.

[106] Competition and Consumer Act 2010 (Cth) sch 2 s 23; Australian Securities and Investments Commission Act 2001 (Cth) s 12BF.

[107] Competition and Consumer Act 2010 (Cth) sch 2 s 21; Australian Securities and Investments Commission Act 2001 (Cth) s 12CB(4).

[108] Jeannie Paterson, ‘The Australian Unfair Contract Terms Law: The Rise of Substantive Unfairness as a Ground for Review of Standard Form Consumer Contracts’ [2009] MelbULawRw 32; (2009) 33(3) Melbourne University Law Review 934, 940.

[109] Kelvin FK Low and Eliza Mik, ‘Lost in Transmission: Unilateral Mistakes in Automated Contracts’ (2020) 136 Law Quarterly Review 563, 568.

[110] Quoine (n 1) [195] (Mance IJ).

[111] Treasury Laws Amendment (Measures for a later sitting) Bill 2021: Unfair contract terms reforms (Cth).

[112] Treasury Laws Amendment (Measures for a later sitting) Bill 2021: Unfair contract terms reforms (Cth).

[113] Electronic Transactions Act 1999 (Cth).

[114] Allen Kiat Peng, ‘Contract Formation and Mistake in Cyberspace (Again)’ (2021) 33 SAcLJ 692, 693.

[115] Quoine (n 1) [205] (Mance IJ).

[116] Electronic Transactions Act 1999 (Cth) ss 5(1), 15D.

[117] Electronic Transactions Act 1999 (Cth) s 3(c).

[118] Harry Surden, ‘Computable Contracts’ (2012) 46 UC Davis Law Review 629, 689, 694.

[119] Dhanoa Harsimar, ‘Making Mistakes With Machines’ (2021) 37(1) Santa Clara High Technology Law Journal 97, 117.

[120] Ibid.

[121] Quoine (n 1) [195] (Mance IJ).

[122] Quoine (n 1).


AustLII: Copyright Policy | Disclaimers | Privacy Policy | Feedback
URL: http://www.austlii.edu.au/au/journals/UNSWLawJlStuS/2022/32.html