“What is left over if I subtract the fact that my arm goes up from the fact that I raise my arm?” Wittgenstein's question opens the research report I wrote at the end of this project, and it's the question that sat behind everything I did this summer as a Laidlaw Scholar in Prof. Olaf Blanke's Laboratory of Cognitive Neuroscience (LNCO), under the supervision of Dr. Bruno Herbelin and Dr. Qiaoyue Ren.
The question
The sense of agency is the subjective experience of controlling your own actions and their effects — the quiet, usually invisible conviction that when you press a light switch, you are the one who turned the light on. We only really notice it when it fails: when the switch doesn't respond, or the outcome doesn't match what we expected.
Episodic memory is the ability to recall specific past events, not just facts. It's reconstructive rather than a neutral recording, which makes it sensitive to internal states — emotion, expectation, and, we hypothesised, agency itself. If an event is self-caused, we reasoned, it should feel more like it "belongs" to you, strengthening its link to your sense of self and making it more memorable. My role this summer was to build the tool that could test that hypothesis at scale.
Building the experiment
Participants move objects across a screen using a keyboard. In some trials, the object's movement matches their input exactly (congruent); in others, it diverges from what they intended (incongruent) — a controlled way of dialing the sense of agency up or down. After each trial, participants rate how much control they felt over the object's movement. Later, in a second session, they're asked whether they've seen each image before, and where it ended up.
Taking the experiment online — rather than lab-only — meant it could reach a much wider and more representative pool of participants. My main technical contribution was building that online platform from the ground up, and adding a feature the lab hadn't used before: confidence ratings after each recognition trial, to see whether subjective confidence tracks the same agency effect as raw accuracy.
What the data is starting to show
Full data collection for the confidence-rating protocol is still ongoing, but a related earlier study by Dr. Ren and Killian Raude (EPFL Neuro-X Master's student) gives a preview. Participants reported a higher sense of control in congruent trials, as expected. More interestingly, congruency improved memory for where an object ended up — spatial memory — but not for what the object was — item memory. That asymmetry suggests agency strengthens the binding between an action and its consequences specifically, rather than boosting memory indiscriminately.
It's a small but real result, and it points toward something with genuine clinical relevance: hypoagentic states — the diminished or distorted sense of control seen in schizophrenia and Alzheimer's disease — could be connected to the memory deficits that accompany them. If agency really is a retrieval cue, strengthening it might become part of how we support memory in those conditions.
The part I didn't expect to write about
Somewhere in the middle of the summer, I lost my own sense of agency over the work. I leaned too hard on AI tools to write code I didn't fully understand conceptually, got stuck debugging things I couldn't reason about, and stopped feeling the sense of accomplishment I normally get from intellectual work. My attention drifted. It's a strange thing to notice you're studying agency while feeling conspicuously short of it.
With help from one of my supervisors, I deliberately slowed down and retook control of the implementation — writing more of the logic myself, understanding it before shipping it. The work got harder before it got easier, but the sense of ownership came back, and so did the satisfaction. It turned out to be the most honest data point I collected all summer, even if it never made it into a figure.
A higher sense of agency enhanced recollection of the spatial location of objects — evidence that agency does shape episodic memory, particularly for the consequences of our actions.
The full report — background, methodology, results, and a longer discussion of the legal and clinical implications of agency, including the Milgram experiment — is available on the Laidlaw Scholars Network.