David V. Pynadath’s research while affiliated with University of Southern California and other places

Successful human-machine teaming often hinges on the ability of eXplainable Artificial Intelligence (XAI) to make an agent’s reasoning transparent to human teammates. Doing so requires that the agent navigate a tradeoff between revealing its reasoning to those teammates without overwhelming them with too much information. This challenge is amplified when a person is teamed with multiple agents. This amplification is not simply linear, due to the increase from 1 to N agents’ worth of reasoning content, but also due to the interdependency among the agents’ reasoning that must be made transparent as well. In this work, we examine the challenges in conveying this interdependency to people teaming with multiple agents. We also propose alternate domain-independent strategies for a team of simulated robots to generate messages about their reasoning to be conveyed to a human teammate. We illustrate these strategies through their implementation in a search-and-rescue simulation testbed.KeywordsRobotsAvatars and Virtual HumansHuman-Robot InteractionExplainable Artificial Intelligence

An implicit expectation of asking users to rate agents, such as an AI decision-aid, is that they will use only relevant information—ask them about an agent’s benevolence, and they should consider whether or not it was kind. Behavioral science, however, suggests that people sometimes use irrelevant information. We identify an instance of this phenomenon, where users who experience better outcomes in a human-agent interaction systematically rated the agent as having better abilities, being more benevolent, and exhibiting greater integrity in a post hoc assessment than users who experienced worse outcomes—which were the result of their own behavior—with the same agent. Our analyses suggest the need for augmentation of models so they account for such biased perceptions as well as mechanisms so that agents can detect and even actively work to correct this and similar biases of users.KeywordsAgent factorsTrait attributionCognitive biasAgent-user interactions

Behavioral scientists have classically documented aversion to algorithmic decision aids, from simple linear models to AI. Sentiment, however, is changing and possibly accelerating AI helper usage. AI assistance is, arguably, most valuable when humans must make complex choices. We argue that classic experimental methods used to study heuristics and biases are insufficient for studying complex choices made with AI helpers. We adapted an experimental paradigm designed for studying complex choices in such contexts. We show that framing and anchoring effects impact how people work with an AI helper and are predictive of choice outcomes. The evidence suggests that some participants, particularly those in a loss frame, put too much faith in the AI helper and experienced worse choice outcomes by doing so. The paradigm also generates computational modeling-friendly data allowing future studies of human-AI decision making.

Optimization of human-AI teams hinges on the AI’s ability to tailor its interaction to individual human teammates. A common hypothesis in adaptive AI research is that minor differences in people’s predisposition to trust can significantly impact their likelihood of complying with recommendations from the AI. Predisposition to trust is often measured with self-report inventories that are administered before interactions. We benchmark a popular measure of this kind against behavioral predictors of compliance. We find that the inventory is a less effective predictor of compliance than the behavioral measures in datasets taken from three previous research projects. This suggests a general property that individual differences in initial behavior are more predictive than differences in self-reported trust attitudes. This result also shows a potential for easily accessible behavioral measures to provide an AI with more accurate models without the use of (often costly) survey instruments.Keywordshuman-robot interactionhuman-computer interactioncompliancetrustinterventiondecision making

To understand how people interact with each other in collaborative settings, especially in situations where individuals know little about their teammates, Multiagent Inverse Reinforcement Learning (MIRL) aims to infer the reward functions guiding the behavior of each individual given trajectories of a team's behavior during task performance. Unlike current MIRL approaches, team members \emph{are not} assumed to know each other's goals a priori, rather they collaborate by adapting to the goals of others perceived by observing their behavior, all while jointly performing a task. To address this problem, we propose a novel approach to MIRL via Theory of Mind (MIRL-ToM). For each agent, we first use ToM reasoning to estimate a posterior distribution over baseline reward profiles given their demonstrated behavior. We then perform MIRL via decentralized equilibrium by employing single-agent Maximum Entropy IRL to infer a reward function for each agent, where we simulate the behavior of other teammates according to the time-varying distribution over profiles. We evaluate our approach in a simulated 2-player search-and-rescue operation where the goal of the agents, playing different roles, is to search for and evacuate victims in the environment. Results show that the choice of baseline profiles is paramount to the recovery of ground-truth rewards, and MIRL-ToM is able to recover the rewards used by agents interacting with either known and unknown teammates.

Optimization of human-AI teams hinges on the AI's ability to tailor its interaction to individual human teammates. A common hypothesis in adaptive AI research is that minor differences in people's predisposition to trust can significantly impact their likelihood of complying with recommendations from the AI. Predisposition to trust is often measured with self-report inventories that are administered before interactions. We benchmark a popular measure of this kind against behavioral predictors of compliance. We find that the inventory is a less effective predictor of compliance than the behavioral measures in datasets taken from three previous research projects. This suggests a general property that individual differences in initial behavior are more predictive than differences in self-reported trust attitudes. This result also shows a potential for easily accessible behavioral measures to provide an AI with more accurate models without the use of (often costly) survey instruments.

An implicit expectation of asking users to rate agents, such as an AI decision-aid, is that they will use only relevant information -- ask them about an agent's benevolence, and they should consider whether or not it was kind. Behavioral science, however, suggests that people sometimes use irrelevant information. We identify an instance of this phenomenon, where users who experience better outcomes in a human-agent interaction systematically rated the agent as having better abilities, being more benevolent, and exhibiting greater integrity in a post hoc assessment than users who experienced worse outcome -- which were the result of their own behavior -- with the same agent. Our analyses suggest the need for augmentation of models so that they account for such biased perceptions as well as mechanisms so that agents can detect and even actively work to correct this and similar biases of users.

Behavioral scientists have classically documented aversion to algorithmic decision aids, from simple linear models to AI. Sentiment, however, is changing and possibly accelerating AI helper usage. AI assistance is, arguably, most valuable when humans must make complex choices. We argue that classic experimental methods used to study heuristics and biases are insufficient for studying complex choices made with AI helpers. We adapted an experimental paradigm designed for studying complex choices in such contexts. We show that framing and anchoring effects impact how people work with an AI helper and are predictive of choice outcomes. The evidence suggests that some participants, particularly those in a loss frame, put too much faith in the AI helper and experienced worse choice outcomes by doing so. The paradigm also generates computational modeling-friendly data allowing future studies of human-AI decision making.