Machine learning models to understand consumer preferences

Adam Hornsby
Industrial Fellow 2018

Adam Hornsby
dunnhumby and University College London

Adam aims to develop a cognitive computational model that describes how people make subjective choices in complex, real-world domains, such as supermarkets. To do this, he will combine insights from model simulations, controlled experiments and analyses of big datasets of real consumer purchasing.  One potential application of Adam’s work will be to create a recommendation engine that can nudge consumers towards healthier purchasing habits.

To do this, Adam is working with a new computational model of subjective decision-making, which shows that as people make subjective choices, they self-reinforce those choices, making them less likely to try new ones. In the supermarket for example, customers may end up liking products because of the fact they have chosen them, which is the opposite to what was previously theorised – that customers occasionally try new products to minimise regret of choosing inferior options. Importantly, Adam’s work suggests that customers form attachments to specific attributes of products, such as packaging or taste, rather than the product itself. This “coherency maximisation” model explains how people develop preferences for products over time and underpins his recommendation system.

"Adam’s work suggests that customers form attachments to specific attributes of products, such as packaging or taste"

Adam is a Senior Data Scientist at dunnhumby and a part-time Experimental Psychology PhD student at UCL.  At dunnhumby, his role is to develop algorithms and software that helps the company’s clients to understand their customers better, helping them to increase profitability and improve customer experiences.

He was the highest scoring psychology undergraduate at UCL, with his final year dissertation focusing on medical decision making. He also won Best New Talent at the DataIQ Awards 2018.

Press

University College London
The Engineer