ORIGINS: grounding artificial intelligence in the origins of human behaviour

Changed on 04/12/2020
In 2019, Inria launched Exploratory Actions, an initiative aimed at showcasing people coming up with new challenges. Say, for instance, that you're exploring theories surrounding the evolution of human behaviour, looking for its founding principles in order to incorporate them into artificial intelligence algorithms. This is the challenge Clément Moulin-Frier and Eleni Nisioti, members of the Flowers project team, have been working on. Let’s take a closer look at this exciting new project, called ORIGINS.
paysage de savane avec des robots
©Images Pixabay_DJSuderman_D1TheOne /Montage L.Chevillot

An exploratory action

In 2020 Clément Moulin-Frier, who has been a researcher within the Flowers project team since 2019, was awarded support from the institute through the “Exploratory Actions” initiative. This initiative was set up to provide researchers with the resources needed for scientific risk-taking, the goal being to boost the dynamism of our research by taking a fresh look at existing research subjects in addition to helping new ones to emerge. However, the proposal Moulin-Frier made was in a particularly competitive field: artificial intelligence. To help him with his challenge, he was able to enlist the support of Eleni Nisioti, a postdoctoral researcher within the team, through the Exploratory Actions initiative.

Finding the right profile

Recruitment began in July 2020, seeking out someone with exactly the right interdisciplinary expertise in areas that tend to have little to do with each other - which sounds like enough of a challenge on its own. In the end however, applications were received from people with solid foundations in many of the desired fields, making them ideal for the project. Quite unexpectedly, the hardest part of the recruitment phase was deciding between what each individual candidate was capable of bringing. Eventually, it was Eleni Nisioti, who was completing her PhD in the UK at the University of Essex, who was chosen to join Flowers for this project. The fact that her arrival on 1st November 2020 coincided with a new national lockdown to counter the spread of the coronavirus did nothing to dampen her enthusiasm for getting started on this innovative project.

Artificial intelligence and human behavioural ecology: an unusual combination

In many ways, this research project is quite unique. Clément and Eleni’s aim is to determine whether the acquisition of complex behaviour by AI systems can be improved by modelling the ecological conditions that influenced human evolution.

Within artificial intelligence today, the focus is almost exclusively on developing algorithms that will enable machines to learn how to solve certain tasks through experience. However, in the living world, and particularly in humans, the dynamics of the environment we have evolved in have had a strong influence on the acquisition of complex behaviour (such as the use of tools or language, for example). There is a whole research field, Human Behavioural Ecology, devoted to exploring how the behaviours which seem to characterise humans as a species can be considered adaptive responses to major changes that have occurred in our environment over the course of our evolution. However, for each individual behaviour, there are a number of different theories regarding the ecological conditions which led to their emergence. Some of these are complementary, but others are contradictory.

This project will primarily involve delving into the vast field of Human Behavioural Ecology in order to extract principles that could be used to design new artificial intelligence algorithms. We will do this by incorporating them into processes which may have guided the emergence of complex cognitive capacities over the course of human evolution explains Clément Moulin-Frier.

The “ORIGINS - rooting artificial intelligence in the origins of human behaviour” exploratory action will thus lead to new interactions between two fields of research that rarely come into contact with each other. Human Behavioural Ecology has of course benefited from research into computer science (modelling, simulation, etc.), but taking this discipline as a central focus in artificial intelligence is entirely new. The long-term objective is for AI to be firmly rooted in our knowledge of the history of human evolution.

This programme is a reminder that humans are very much central to so-called artificial intelligence!

Un programme qui rappelle que l’humain est bien au cœur d’une intelligence dite artificielle !

Clément Moulin-Frier's profile

Portrait de Clément Moulin-Frier

Clément Moulin-Frier graduated with a PhD in the Engineering of Cognition, Interaction, Learning and Creation from the University of Grenoble in 2011. The focus of his research has primarily been on the emergence of social behaviours in populations of organic or artificial agents, employing an approach centred around computer simulation and modelling.  In 2009, he spent time as a guest researcher at Michel Arbib’s laboratory at the University of Southern California. 

 After a period spent at the Collège de France in Paris, where he worked on probabilistic optimisation for biped robots, he joined the Flowers project team at the Inria Bordeaux - Sud-Ouest research centre, where he carried out research in the field of developmental robotics between 2011 and 2014. From 2015 to 2017 he worked at the SPECS laboratory in Barcelona (Synthetic, Perceptive, Emotive and Cognitive Systems Laboratory), during which time he took part in major EU projects in social robotics such as WYSIWYD and SocSMCs

From 2017 to 2019 he was a researcher for the American artificial intelligence research company Cogitai (which has since become Sony AI America). He returned to Flowers in October 2019, this time as a permanent research fellow.

Eleni Nisioti's profile

Portrait d'Eleni Nisioti

Originally from Greece, Eleni Nisioti studied for an undergraduate degree in electrical and computer engineering at the Aristotle University of Thessaloniki, with a particular focus on machine learning. She then stayed in the world of research, moving to the University of Essex from where she has just graduated with a PhD in the application of reinforcement learning in communication networks.  She joined the Flowers project team as a postdoctoral researcher on 1st November and will stay there for two years.