Ido Ben Artzi

Controlling outcome-irrelevant learning in complex environments

Project description

Children often learn from direct experience. They playfully form mental representations of the external world by acting and observing the associated outcomes. Current educational studies suggest clear benefits for learning from experience, compared to passively learning from instructive and frontal teaching. However, it is not clear whether learning from experience is indeed beneficial to all individuals in a similar manner. Imagine a child solving a Jigsaw puzzle in an active-learning session in class. The learning process occurs when after succeeding in putting together a piece, the child assigns a positive value to the preceding action. However, the action may be complex, and will therefore require the child to establish a mental representation of the environment. Thus, some children may falsely assign value to a motor or visual feature of the action which did not cause the rewarding outcome.

Recent research from our lab suggests that when learning from experience, some individuals tend to assign value to noise, and learn from incidental events that do not reflect the true environment structure (i.e., outcome-irrelevant learning). Thus, for these individuals, learning from direct experience might be confusing and counterproductive. By deriving conclusions from basic science about the phenomenon, we aim to formulate guidelines on how teachers can help children make better use of learning from experience in class.

Our reinforcement learning computational models allow us to derive estimates for subjective values that learners assign to features of their actions. This enables us to distinguish between individuals that assign value to outcome-irrelevant features of their actions and those that do not. Then, one possible intervention relates to the use of psycho-educational instructions. Specifically, we could teach students about the random occurrence of events, and occupy them with ways to inhibit learning from noise. Furthermore, our extensive training research will enable us to evaluate whether some students need to occasionally refresh their memory regarding what is causing the reward. Finally, our experimental paradigm may shed light on how motivational interventions can mitigate this learning bias.

About me

After completing my BSc in Psychology and Biology with an emphasis in Neuroscience, I joined Dr. Nitzan Shahar’s Lab. In the lab, we focus on the computational modeling of human reinforcement learning. Furthermore, due to Nitzan’s background as a clinical psychologist, we aim to achieve a deep theoretical understanding of cognitive mechanisms that may underlie clinical symptoms.

Personally, my interest in learning and decision-making starts from my professional career as a chess International Master. Growing up in this world, I was fascinated by aspects such as pattern recognition, planning, and analytical thinking. Moreover, experiencing first-hand how these can be affected by the stress accompanying high-level tournaments made me want to pursue research in this field. Over the last 10 years, I also had the chance to teach and guide young chess players as they learn and develop through the ranks.

Other than that, I am interested in the translation of scientific knowledge into products that could help solve daily challenges. In that regard, I have volunteered for the last years in a non-profit organization called Brainstorm IL which aims to connect academia to the Neuro-tech industry and by that to develop a vibrant Israeli ecosystem in our field.