Timezone: »

 
When Humans Aren’t Optimal: Robots that Collaborate with Risk-Aware Humans
Minae Kwon · Erdem Biyik · Aditi Talati · Karan Bhasin · Dylan Losey · Dorsa Sadigh

@
in Cooperative AI »
Event URL: https://openreview.net/forum?id=aa_HSq-q-0n »
In order to collaborate safely and efficiently, AI agents need to anticipate how their human partners will behave. Some of today’s agents model humans as if they were also agents, and assume users are always optimal. Other agents account for human limitations, and relax this assumption so that the human is noisily rational. Both of these models make sense when the human receives deterministic rewards: i.e., gaining either $100 or $130 with certainty. But in real-world scenarios, rewards are rarely deterministic. Instead, we must make choices subject to risk and uncertainty— and in these settings, evidence suggests humans exhibit a cognitive bias towards suboptimal behavior [1]. For example, when deciding between gaining $100 with certainty or $130 only 80% of the time, people tend to make the risk-averse choice— even though it leads to a lower expected gain! In this paper, we adopt a well-known Risk-Aware human model from behavioral economics called Cumulative Prospect Theory and enable agents to leverage this model during human-agent interaction. In our user studies, we offer supporting evidence that the Risk-Aware model more accurately predicts suboptimal human behavior. We find that this increased modeling accuracy results in safer and more efficient human-agent collaboration. Overall, we extend existing rational human models so that collaborative agents can anticipate and plan around suboptimal human behavior during human-agent interaction.

Author Information

Minae Kwon (Stanford University)
Erdem Biyik (Stanford University)

Erdem Biyik is a PhD candidate in Electrical Engineering at Stanford University. He is working on AI for Robotics in Intelligent and Interactive Autonomous Systems Group (ILIAD), and advised by Prof. Dorsa Sadigh. His research interests are: machine learning, artificial intelligence (AI), and their applications for human-robot interaction and multi-agent systems. He also works on AI and optimization for autonomous driving and traffic management. Before coming to Stanford, Erdem was an undergraduate student in the Department of Electrical and Electronics Engineering at Bilkent University, where he worked in Imaging and Computational Neuroscience Laboratory (ICON Lab) in National Magnetic Resonance Research Center under the supervision of Prof. Tolga Çukur with a focus on compressed sensing reconstructions, coil compression, and bSSFP banding suppression in MRI. He also worked on generalized approximate message passing algorithms as an intern in Prof. Rudiger Urbanke's Communication Theory Laboratory (LTHC) in EPFL, under the supervision of Dr. Jean Barbier, for a summer.

Aditi Talati
Karan Bhasin
Dylan Losey (Virginia Tech)
Dorsa Sadigh (Stanford)

More from the Same Authors