Forget modeling every belief and goal! What if we represented people as following simple scripts instead (i.e "cross the crosswalk")?

Our new paper shows AI which models others’ minds as Python code 💻 can quickly and accurately predict human behavior!

shorturl.at/siUYI%F0%9F%...

I like how this viewpoint emphasizes how AI can have both profound, transformative (and potentially risky) effects on society while not fixating on a catastrophic scenario -- it definitely makes it easier, as the article states, "to grapple with the weirdness and disruption that is happening today."

Companies are pouring billions into humanoid #robotics. In a new @spectrum.ieee.org ‬ op-ed, #UWAllen professor @mayacakmak.bsky.social wonders whether people would really welcome such robots into their homes — and looks forward to taking one for a spin for science. spectrum.ieee.org/home-humanoi...

While it's important to note that our findings may not generalize to the full population (n=21 vignette survey), they have major implications for companies rushing to deploy autonomous household robots with remote teleoperators. It challenges the industry assumption that "more autonomous = better."

3) End user programmed robots that learn from the user’s demonstrations/instructions are a promising paradigm for autonomous robots that give the user higher control over the robot’s intentions. This translated into a significantly higher sense of agency rating from the participants.

2) Preferences for robot autonomy and amount of user control vary with the context & its risk. Users preferred to have full control over the robot in high risk scenarios, and preferred the opposite in low risk scenarios. Privacy and social concerns enhanced their risk rating & desire for control.

✏️ Key findings:
1) Third-party involvement in robot programming or control had a significantly larger negative effect on the participants' sense of agency than whether the robot acted autonomously.

But the big question is: do people want this? How does it affect their sense of agency (aka their feeling of control) over their environment? What factors determine their preferences?

Companies like 1X and Weave have announced their plans to deploy assistive robots. These robots will act autonomously, with third party teleoperators ready to intervene in case of failure. This mode of operation is necessary, as robot behaviors are not yet robust enough for households.

I’m presenting our paper “Preserving Sense of Agency: User Preferences for Robot Autonomy and User Control across Household Tasks” at IEEE RO-MAN 2025!

We challenge a core assumption in robotics: do users want autonomous household robots with third party teleoperators? 🧵 arxiv.org/abs/2506.19202

Still catching up on my notes after my first #cogsci2025, but I'm so grateful for all the conversations and new friends and connections! I presented my poster "When Empowerment Disempowers" -- if we didn't get the chance to chat or you would like to chat more, please reach out!