Aalto and ELLIS Institute researchers to build AI assistants that learn from minimal user data — as project targets "cold start" problem in personalisation
Business Finland-funded RealYou AI project combines cognitive science with machine learning. The aim is for AI systems to learn to understand individual users from a small amount of interaction, without the kind of large-scale behavioural tracking currently in use.
Text by Martti Asikainen, 8.7.2026 | Photo by Adobe Stock Photos
Researchers at Aalto University and ELLIS Institute Finland have launched a project to develop AI-based decision assistants. ELLIS Institute Finland is a European AI research institute based in Helsinki.
The new assistants are intended to learn a user’s preferences from just a handful of interactions. This marks a considerable leap from current recommendation systems, which require large volumes of data to function.
The RealYou AI project is funded by Business Finland. This is confirmed by announcements published by both institutions on 1 July. Neither announcement discloses the size of the funding or the project’s duration, though this is likely to become clear in the coming months.
Help with a widespread problem?
Most existing digital assistants and recommendation systems rely on accumulated interaction data. This approach performs poorly in two situations: when a user is new or uses a service infrequently, and when preferences are complex or change over time.
In the industry, this is often referred to as the “cold start problem” — a situation in which a system has too little information about a new user to make relevant, content-based recommendations. RealYou AI is built around what is known as cognitive machine learning.
This approach models why people make decisions and how those decisions vary between individuals, rather than relying solely on what large datasets reveal on average.
According to the joint announcement, the project will investigate technology that learns user preferences from a small number of interactions and adapts continuously as circumstances change. It also aims to make AI-generated recommendations more transparent while reducing the amount of personal data such systems need to collect.
"Significant value left on the table"
According to the institutions, potential applications include e-commerce, financial services, healthcare, education and professional decision support, though neither named specific industry partners or planned deployments.
“Personalisation is becoming one of the defining challenges and opportunities in artificial intelligence,” says Andrew Howes, visiting professor at ELLIS Institute Finland and professor at the University of Exeter.
According to Howes, generic AI experiences leave organisations with “significant value on the table”, and future systems will need to understand individuals rather than populations.
The research team is led by Samuel Kaski, director of ELLIS Institute Finland and professor at Aalto University. Also involved are Howes, doctoral researcher Alex Kortia, and Aalto University lecturer Antti Lähtevänoja. The project sits within Aalto University’s Department of Computer Science and is being carried out in collaboration with the ELLIS Institute.
Business Finland has funded a number of AI research projects in Finland in recent years. The announcement does not, however, specify how RealYou AI’s funding compares with other awards, or on what basis the project was selected for funding.
