Data Study Group - May 2024

Introduction

Stage 1: Precursor Stage (part-time, online)

• The Precursor Stage runs from Monday 13 May – Friday 17 May 2024, in the lead up to the Event Stage.
• The maximum time commitment is 2.5 hours a day.
• This includes online workshops, presentations and team building which will prepare participants for the Event Stage.

Stage 2: Event Stage (full-time, in person)

• The Event Stage runs from Monday 20 May - Friday 24 May 2024, and is held at The Alan Turing Institute (British Library, 96 Euston Road, London NW1 2DB).
• Group work begins and continues throughout.

Applicants should be able to commit to the duration of the event. The Alan Turing Institute is committed to supporting individual circumstances, please do not hesitate to email [email protected] to discuss any reasonable adjustments. 

 

Challenges 

Kunato

Utilising machine learning and casual statistics to generate trust index for news publishers, and derive the relation between a publisher’s trustworthiness and its audience engagement/revenue.

Kunato is an AI economics research and deployment company that uses cutting-edge deep learning algorithms to dynamically price every piece of internet content and create a new market for information commerce.

Due to the abundance of fake content, the digital landscape is becoming increasingly complex: understanding how a publisher's trust index relates to user engagement and revenue is crucial. This DSG challenge can significantly improve transparency among publishers, combat the spread of fake news and help create a more reliable and trustworthy online content environment.

The challenge is to develop a framework based on machine learning that can derive the trust vector (a culmination of ML-derived factors representing credibility) for a publisher and a causal ML system that can be used to understand and simulate the impact of the same on user engagement and revenue.

Join us in this DSG challenge to enable more informed content valuation by assessing content credibility and taking a step towards a more honest information landscape.
 

Mastercard

Addressing fairness/bias in AI models for financial transactions.

Mastercard aims to evaluate the presence of fairness and bias within their deployed ML models – particularly those employed in financial transactions. Despite a surge in research on AI bias and fairness, translating these insights into practical applications encounters gaps. This data study group challenge provides an opportunity to bridge these gaps by leveraging financial transaction data. Key research questions for this challenge include:
Q1: What novel measures can be developed to ensure responsible AI and fairness in the financial sector, particularly when dealing with multilabel and multi-class models for credit and transactional data?
Q2: How can AI be effectively aligned with diverse contextual normative statements, encompassing fairness, political principles, distributive justice and moral norms?

This DSG challenge not only tackles the technical aspects of AI bias in financial transactions but also dives into the broader socio-technical landscape, contributing to the responsible and fair deployment of AI models in the financial sector.

Theyr

Using explainable AI to improve user trust and adoption in maritime routing.

Autonomous digital systems are helping to immediately reduce emissions across the maritime sector.

T-VOS is a commercial shipping voyage routing system that has been shown to reduce a ship’s fuel consumption by more than 5 per cent above current industry standard solutions.

However, ship captains may decide on alternative routes based on historical perceptions – particularly in scenarios where they cannot understand why the software has selected a specific route.  Communicating this information so that it is easily understood by the end users helps increase the uptake of autonomous systems.

This DSG challenge will investigate how to interpret the decision made by these systems and communicate to the end-users, while also considering their preferences. The objective is to improve user confidence in the commercial software T-VOS, which will help to further decarbonise the shipping industry.

We will do this by developing a tool to interpret the decision made by the voyage optimisation systems and communicate to the end-users the reason a certain route could be beneficial. Specifically, the objectives of this DSG challenge are to:

• Isolate the features that cause the differences in performance of routes generated by the tool.
• Understand how the end-user will interpret these features most easily.
• Develop a method to communicate the relevant features to the user.
• Develop new guidance on design of XAI/M for effective human-computer communication, as well as on supporting explainability of these technologies to end-users – helping to build user confidence, trust and connection.