Autonomous AI systems will be entering human society in the near future to provide services and work alongside humans. For those accepted trusted, users should able understand reasoning process of system, i.e. system transparent. System transparency enables humans form coherent explanations system's decisions actions. Transparency is important not only for user trust, but also software debugging certification. In recent years, Deep Neural Networks have made great advances multiple...

In this paper, we study human trust and its computational models in supervisory control of swarm robots with varied levels autonomy (LOA) a target foraging task. We implement three LOAs: manual, mixed-initiative (MI), fully autonomous LOA. While the MI LOA is controlled by operator an search algorithm collaboratively, swarms manual LOAs are directed algorithm, respectively. From user studies, find that humans tend to make their decisions based on physical characteristics rather than...

In this paper, we study trust-related human factors in supervisory control of swarm robots with varied levels autonomy (LOA) a target foraging task. We compare three LOAs: manual, mixed-initiative (MI), and fully autonomous LOA. the manual LOA, operator chooses headings for flocking swarm, issuing new as needed. is redirected automatically by changing using search algorithm. if performance declines, switched from to or human. The result work extends current knowledge on control....

As domestic service robots become more common and widespread, they must be programmed to efficiently accomplish tasks while aligning their actions with relevant norms. The first step equip normative reasoning competence is understanding the norms that people apply behavior of in specific social contexts. To end, we conducted an online survey Chinese United States participants which asked them select preferred action a robot should take number scenarios. paper makes multiple contributions....

The ability to collaborate with previously unseen human teammates is crucial for artificial agents be effective in human-agent teams (HATs). Due individual differences and complex team dynamics, it hard develop a single agent policy match all potential teammates. In this article, we study both human-human HAT dyadic cooperative task, Team Space Fortress. Results show that the performance influenced by players' skill level their different adopting complementary policies. Based on results,...

Neural agents trained in reinforcement learning settings can learn to communicate among themselves via discrete tokens, accomplishing as a team what would be unable do alone. However, the current standard of using one-hot vectors communication tokens prevents from acquiring more desirable aspects such zero-shot understanding. Inspired by word embedding techniques natural language processing, we propose neural agent architectures that enables them derived learned, continuous space. We show...

The ability to make inferences about other’s mental state is referred as having a Theory of Mind (ToM). Such the foundation many human social interactions such empathy, teamwork, and communication. As intelligent agents being involved in diverse human-agent teams, they are also expected be socially become effective teammates. To provide feasible baseline for future agents, this paper presents experimental study on process ToM reference. Human observers’ compared with participants’ verbally...

With the development of AI technology, intelligent agents are expected to team with humans and adapt their teammates in changing environments, as effective human members would do. As an initial step towards adaptive agents, present study examined individual's actions a cooperative task. By analyzing performance when participants paired different partners, we were able identify adaptations isolate individual contributions performance. It is shown that determined by factors at both levels....

Theory of Mind (ToM) refers to the ability make inferences about other's mental states. Such is fundamental for human social activities such as empathy, teamwork, and communication. As intelligent agents come be involved in diverse human-agent teams, they will also expected socially order become effective teammates. In this paper, we describe a computational ToM model which observes team behaviors infers their states urban search rescue (US&R) task. Our modular approximates inference by...

Control of robotic swarms through control over a leader(s) has become the dominant approach to supervisory these largely autonomous systems. Resilience in face attrition is one primary advantages attributed yet presence makes them vulnerable decapitation. Algorithms which allow swarm hide its leader are promising solution. We present novel neural networks, NNs, trained graph network, GNN, replace conventional controllers making more amenable training. Swarms and an adversary intent finding...

Leader-follower navigation is a popular class of multi-robot algorithms where leader robot leads the follower robots in team. The has specialized capabilities or mission critical information (e.g. goal location) that followers lack, and this makes crucial for mission's success. However, also vulnerability -an external adversary who wishes to sabotage team's can simply harm whole would be compromised. Since motion generated by traditional leader-follower reveal identity leader, we propose...

Detecting and recognizing targets from unmanned aerial vehicle (UAV) video is a common task for operators controlling, supervising, or monitoring UAVs. While everywhere rely on visual interpretation of UAV videos to perform their tasks, we are unaware prior work quantifying the difficulty these subject this domain particularly at extremes. target detection recognition may pose few problems low slow flying UAVs, more demanding applications requiring rapid surveillance larger regions approach...

Multi-Agent Reinforcement Learning (MARL) methods have shown promise in enabling agents to learn a shared communication protocol from scratch and accomplish challenging team tasks. However, the learned language is usually not interpretable humans or other co-trained together, limiting its applicability ad-hoc teamwork scenarios. In this work, we propose novel computational pipeline that aligns space between MARL with an embedding of human natural by grounding agent communications on...

This work studied human teamwork with a concentration on the influence of team synchronization and in- dividual differences performance. Human participants were paired to complete collaborative tasks in simulated game environment, which they assigned roles corresponding responsibilities. Cross- correlation analysis was employed quantify degree time-lag between two teammates’ collective actions. Results showed that performance is determined by factors at both individual levels. We found...

In a search and rescue scenario, rescuers may have different knowledge of the environment strategies for exploration. Understanding what is inside rescuer's mind will enable an observer agent to proactively assist them with critical information that can help perform their task efficiently. To this end, we propose build models based on trajectory observations predict strategies. our efforts model mind, begin simple simulated in Minecraft human participants. We formulate neural sequence triage...

The interaction between swarm robots and human operators is significantly different from the traditional humanrobot due to unique characteristics of system, such as high cognitive complexity difficulties in state estimation. In this paper, we concentrated on method conveying input operator swarm. Previous research has shown that control through switching behaviors offers greatest flexibility but particularly difficult for operators. A recently developed finding optimal sequences composing...

Autonomous AI systems will be entering human society in the near future to provide services and work alongside humans. For those accepted trusted, users should able understand reasoning process of system, i.e. system transparent. System transparency enables humans form coherent explanations system's decisions actions. Transparency is important not only for user trust, but also software debugging certification. In recent years, Deep Neural Networks have made great advances multiple...

Trust is an important factor in the interaction between humans and automation that can mediate reliance of human operators. In this work, we evaluate a computational model trust on swarm systems based Sheridan (2019)’s modified Kalman estimation using existing experiment data (Nam, Li, Lewis, & Sycara, 2018). Results show our Filter outperforms state art alternatives including dynamic Bayesian networks inverse reinforcement learning. This work novel that: 1) The estimator first...