When To Seek Help: Trust-Aware Assistance Seeking in Human-Supervised Autonomy

TL;DR

This work tackles trust dynamics in human-supervised autonomy by modeling trust as a latent state $T$ driven by an Input-Output Hidden Markov Model and integrating it into a Partially Observable Markov Decision Process to derive a trust-aware assistance-seeking policy. The approach is validated through two human-in-the-loop experiments using a mobile manipulator tasked with object collection, revealing that requesting human help in high-complexity scenarios can repair and enhance trust, while avoiding unnecessary interruptions. The computed policy, obtained via value iteration with $\\gamma=0.99$, outperforms a trust-agnostic baseline, achieving higher cumulative rewards and greater post-experiment trust (e.g., threshold $b^T>0.73$ in $C^H$). The results establish a principled, model-based framework for calibrating trust in HRI and demonstrate the practical impact of trust-aware decision-making on team performance and user trust.

Abstract

Our goal is to model and experimentally assess trust evolution to predict future beliefs and behaviors of human-robot teams in dynamic environments. Research suggests that maintaining trust among team members in a human-robot team is vital for successful team performance. Research suggests that trust is a multi-dimensional and latent entity that relates to past experiences and future actions in a complex manner. Employing a human-robot collaborative task, we design an optimal assistance-seeking strategy for the robot using a POMDP framework. In the task, the human supervises an autonomous mobile manipulator collecting objects in an environment. The supervisor's task is to ensure that the robot safely executes its task. The robot can either choose to attempt to collect the object or seek human assistance. The human supervisor actively monitors the robot's activities, offering assistance upon request, and intervening if they perceive the robot may fail. In this setting, human trust is the hidden state, and the primary objective is to optimize team performance. We execute two sets of human-robot interaction experiments. The data from the first experiment are used to estimate POMDP parameters, which are used to compute an optimal assistance-seeking policy evaluated in the second experiment. The estimated POMDP parameters reveal that, for most participants, human intervention is more probable when trust is low, particularly in high-complexity tasks. Our estimates suggest that the robot's action of asking for assistance in high-complexity tasks can positively impact human trust. Our experimental results show that the proposed trust-aware policy is better than an optimal trust-agnostic policy. By comparing model estimates of human trust, obtained using only behavioral data, with the collected self-reported trust values, we show that model estimates are isomorphic to self-reported responses.

Figures (15)

• Figure 1: Experimental Setup. The top figure shows the world view of the environment available to the human supervisor. The bottom left and middle figures show the local view and end-effector view used by humans during teleoperation. The bottom right figure shows the operation status. The setup uses ROS-Gazebo Gazebo and resources available in downs2022googleROSMoMa.
• Figure 2: Collaborative object collection setup. The human behavioral model with trust as a hidden state yields the likelihood of human intervention and is used by the robot to design an optimal policy in a POMDP setting.
• Figure 3: IOHMM-based dynamic human behavioral model, where the human action $o_t$ is modulated by the hidden trust dynamics $T_t$
• Figure 4: Estimated observation probabilities $\mathbb{P}(o_t| T, C, a)$
• Figure 5: Estimated trust state transition probabilities $\mathbb{P}(T_{t+1}|T_t, E_{t+1}, C_t, a_t)$ for low environment complexity $C_t=C^L$