At the AI's discretion

Where a contract confers a discretion on one party that materially affects the rights of its counterparty,  the discretion must be exercised rationally. The Supreme Court held in Braganza v BP Shipping Ltd [2015] UKSC 17 that exercising a discretion rationally involves (i) taking the right things (and only the right things) into account, and (ii) avoiding a decision that no reasonable decision-maker could have reached. In this episode, we explore how those principles might operate in the context of a discretion exercised automatically by a machine learning algorithm. We do so in the context of a fraud detection algorithm and an online farmers' market somewhere in East Anglia.

Further reading:

• This episode was inspired by Tom Whittaker's thought-provoking article on the case of TF Global Markets (UK) (trading as ThinkMarkets)) v Financial Ombudsman Service Limited [2020] EWHC 3178 (Admin). The article may be found here: https://www.lexology.com/library/detail.aspx?g=ad5569ea-af1a-4040-b596-a6a29b3c73b0
• Supreme Court decision in Braganza v BP Shipping Ltd [2015] UKSC 17: https://www.supremecourt.uk/cases/uksc-2013-0099.html
• Anyone with any doubts as to the prevalence of AI-based fraud detection systems might like to do this: https://letmegooglethat.com/?q=ai+fraud+detection. There is no problem in principle with using such tools. The issue (in a contractual context) is how their outputs are translated into discretionary decisions.