An official website of the United States government
Much prior work on creating social agents that assist users relies on preconceived assumptions of what it means to be helpful. For example, it is common to assume that a helpful agent just assists with achieving a user’s objective. However, as assistive agents become more widespread, human-agent interactions may be more ad-hoc, providing opportunities for unexpected agent assistance. How would this affect human notions of an agent’s helpfulness? To investigate this question, we conducted an exploratory study (N=186) where participants interacted with agents displaying unexpected, assistive behaviors in a Space Invaders game and we studied factors that may influence perceived helpfulness in these interactions. Our results challenge the idea that human perceptions of the helpfulness of unexpected agent assistance can be derived from a universal, objective definition of help. Also, humans will reciprocate unexpected assistance, but might not always consider that they are in fact helping an agent. Based on our findings, we recommend considering personalization and adaptation when designing future assistive behaviors for prosocial agents that may try to help users in unexpected situations.
more »
« less
Stop Copying Me: Evaluating nonverbal mimicry in embodied motivational agents
Motivational agents are virtual agents that seek to motivate users by providing feedback and guidance. Prior work has shown how certain factors of an agent, such as the type of feedback given or the agent’s appearance, can influence user motivation when completing tasks. However, it is not known how nonverbal mirroring affects an agent’s ability to motivate users. Specifically, would an agent that mirrors be more motivating than an agent that does not? Would an agent trained on real human behaviors be better? We conducted a within-subjects study asking 30 participants to play a “find-the-hidden-object” game while interacting with a motivational agent that would provide hints and feedback on the user’s performance. We created three agents: a Control agent that did not respond to the user’s movements, a simple Mimic agent that mirrored the user’s movements on a delay, and a Complex agent that used a machine-learned behavior model. We asked participants to complete a questionnaire asking them to rate their levels of motivation and perceptions of the agent and its feedback. Our results showed that the Mimic agent was more motivating than the Control agent and more helpful than the Complex agent. We also found that when participants became aware of the mimicking behavior, it can feel weird or creepy; therefore, it is important to consider the detection of mimicry when designing virtual agents.
more »
« less
- Award ID(s):
- 1750840
- PAR ID:
- 10534485
- Publisher / Repository:
- ACM
- Date Published:
- ISBN:
- 9781450399944
- Page Range / eLocation ID:
- 1 to 4
- Format(s):
- Medium: X
- Location:
- Würzburg Germany
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Cromarty, L.; Shirwaiker, R; Wang, P. (Ed.)When natural disasters occur, unaffiliated volunteers are inspired to help within their community and are known as spontaneous volunteers (SVs). Our research seeks to understand SV convergence as it relates to individual SV motivation, engagement, and decision making. We developed an agent-based model in AnyLogic to simulate the decision-making process of potential SV agents during disasters and how it affects volunteer response. Internal motivation is indicative of an agent’s willingness to volunteer, which was modeled by the Theory of Planned Behavior (TPB). We alter motivational factors to assess how they impact SV participation. We examine SV engagement by exploring targeted versus random messaging from volunteer sites to agents. Agents select volunteer sites based on information sharing policies common within the social network literature. Results show that a site choice decision based on connections (friends) negatively influenced demand completion. Alternatively, having pre-existing confidence in abilities and non-targeted volunteer site to agent messaging positively influences the number of participating SVs and therefore decreases demand most significantly over a 30-day period.more » « less
-
Laypeople often believe that God punishes transgressions; however, their inferences about God’s punishment motives remain unclear. We addressed this topic by asking laypeople to indicate why God punishes. We also examined participants’ inferences about why humans punish to contribute to scholarly conversations regarding the extent to which people may anthropomorphize God’s mind. In Studies 1A to 1C, participants viewed God as less retributive than humans. In Study 2, participants expected God (vs. humans) to view humans’ true selves more positively; this difference mediated participants’ views of God as less retributive than humans. Study 3 manipulated agents’ views of humans’ true selves and examined how such information influenced each agent’s perceived motives. Participants viewed a given agent as less retributive when that agent regarded the true self as good (versus bad). These findings extend scholarship on lay theories of punishment motives and highlight links between religious and moral cognition.more » « less
-
Prior work has shown that embodiment can benefit virtual agents, such as increasing rapport and conveying non-verbal information. However, it is unclear if users prefer an embodied to a speech-only agent for augmented reality (AR) headsets that are designed to assist users in completing real-world tasks. We conducted a study to examine users' perceptions and behaviors when interacting with virtual agents in AR. We asked 24 adults to wear the Microsoft HoloLens and find objects in a hidden object game while interacting with an agent that would offer assistance. We presented participants with four different agents: voice-only, non-human, full-size embodied, and a miniature embodied agent. Overall, users preferred the miniature embodied agent due to the novelty of his size and reduced uncanniness as opposed to the larger agent. From our results, we draw conclusions about how agent representation matters and derive guidelines on designing agents for AR headsets.more » « less
-
As robots become pervasive in human environments, it is important to enable users to effectively convey new skills without programming. Most existing work on Interactive Reinforcement Learning focuses on interpreting and incorporating non-expert human feedback to speed up learning; we aim to design a better representation of the learning agent that is able to elicit more natural and effective communication between the human trainer and the learner, while treating human feedback as discrete communication that depends probabilistically on the trainer’s target policy. This work entails a user study where participants train a virtual agent to accomplish tasks by giving reward and/or punishment in a variety of simulated environments. We present results from 60 participants to show how a learner can ground natural language commands and adapt its action execution speed to learn more efficiently from human trainers. The agent’s action execution speed can be successfully modulated to encourage more explicit feedback from a human trainer in areas of the state space where there is high uncertainty. Our results show that our novel adaptive speed agent dominates different fixed speed agents on several measures of performance. Additionally, we investigate the impact of instructions on user performance and user preference in training conditions.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1145/3570945.3607322
