This paper presents a taxonomy of explainability in human-agent systems. We consider fundamental ... more This paper presents a taxonomy of explainability in human-agent systems. We consider fundamental questions about the Why, Who, What, When and How of explainability. First, we define explainability, and its relationship to the related terms of interpretability, transparency, explicitness, and faithfulness. These definitions allow us to answer why explainability is needed in the system, whom it is geared to and what explanations can be generated to meet this need. We then consider when the user should be presented with this information. Last, we consider how objective and subjective measures can be used to evaluate the entire system. This last question is the most encompassing as it will need to evaluate all other issues regarding explainability.
This paper presents objective metrics for how explainable artificial intelligence (XAI) can be qu... more This paper presents objective metrics for how explainable artificial intelligence (XAI) can be quantified. Through an overview of current trends, we show that many explanations are generated post-hoc and independent of the agent's logical process, which in turn creates explanations with limited meaning as they lack transparency and fidelity. While user studies are a known basis for evaluating XAI, studies that do not consider objective metrics for evaluating XAI may have limited meaning and may suffer from confirmation bias, particularly if they use low fidelity explanations unnecessarily. To avoid this issue, this paper suggests a paradigm shift in evaluating XAI that focuses on metrics that quantify the explanation itself and its appropriateness given the XAI goal. We suggest four such metrics based on performance differences, D, between the explanation's logic and the agent's actual performance, the number of rules, R, outputted by the explanation, the number of features, F , used to generate that explanation, and the stability, S, of the explanation. We believe that user studies that focus on these metrics in their evaluations are inherently more valid and should be integrated in future XAI research.
Adaptive Agents and Multi-Agents Systems, May 2, 2011
Creating agents that properly simulate and interact with people is critical for many applications... more Creating agents that properly simulate and interact with people is critical for many applications. Towards creating these agents, models are needed that quickly and accurately predict how people behave in a variety of domains and problems. This paper explores how one bounded rationality theory, Aspiration Adaptation Theory (AAT), can be used to aid in this task. We extensively studied two types of problems -a relatively simple optimization problem and two complex negotiation problems. We compared the predictive capabilities of traditional learning methods with those where we added key elements of AAT and other optimal and bounded rationality models. Within the extensive empirical studies we conducted, we found that machine learning models combined with AAT were most effective in quickly and accurately predicting people's behavior.
Adaptive Agents and Multi-Agents Systems, May 5, 2014
Agents that interact with humans are known to benefit from modeling them. Therefore, when designi... more Agents that interact with humans are known to benefit from modeling them. Therefore, when designing agents intended for interaction with automated agents, it is crucial to model the other agents. However, little is known about how to model automated agents and in particular non-expert agents. Are automated agents to be modeled the same way that an agent models humans? Or does a separate model for interacting with automated agents need to be developed? We evaluate automated agent behavior (for non-expert agents) using a game called the Trust-Revenge game, which is known in social science for capturing different human tendencies. The Trust-Revenge game has a unique sub game-perfect equilibrium, however, very rarely do people follow it. We compared the behavior of automated agents to that of human actions in several demographic groups (including a group which is similar but not identical to the designers of the automated agents). We show that differences between automated agents' and humans' behavior are similar to differences between different human cultures.
Internet social networks have become a ubiquitous application allowing people to easily share tex... more Internet social networks have become a ubiquitous application allowing people to easily share text, pictures, and audio and video files. Popular networks include WhatsApp, Facebook, Reddit and LinkedIn. We present an extensive study of the usage of the WhatsApp social network, an Internet messaging application that is quickly replacing SMS messaging. In order to better understand people's use of the network, we provide an analysis of over 6 million messages from over 100 users, with the objective of building demographic prediction models using activity data. We performed extensive statistical and numerical analysis of the data and found significant differences in WhatsApp usage across people of different genders and ages. We also inputted the data into the Weka data mining package and studied models created from decision tree and Bayesian network algorithms. We found that different genders and age demographics had significantly different usage habits in almost all message and gr...
This paper presents an extensive study of the usage of the WhatsApp social network, an Internet m... more This paper presents an extensive study of the usage of the WhatsApp social network, an Internet messaging application that is quickly replacing SMS messaging. It is based on the analysis of over 4 million messages from nearly 100 users that we collected in order to understand people’s use of the network. We believe that this is the first in-depth study of the properties of WhatsApp messages with an emphasis of noting differences across different age and gender demographic groups. It is also the first to use statistical and data analytic tools in this analysis. We found that different genders and age demographics had significantly different usage habits in almost all message and group attributes. These differences facilitate the development of user prediction models based on data mining tools. We illustrate this by developing several prediction models such as for a person’s gender or approximate age. We also noted differences in users’ group behavior. We created group behaviorial mod...
BACKGROUND Social networks on the Internet have become ubiquitous applications that allow people ... more BACKGROUND Social networks on the Internet have become ubiquitous applications that allow people to easily share text, pictures, and audio and video files. Popular networks include WhatsApp, Facebook, Reddit, and LinkedIn.
This paper presents a taxonomy of explainability in Human-Agent Systems. We consider fundamental ... more This paper presents a taxonomy of explainability in Human-Agent Systems. We consider fundamental questions about the Why, Who, What, When and How of explainability. First, we define explainability, and its relationship to the related terms of interpretability, transparency, explicitness, and faithfulness. These definitions allow us to answer why explainability is needed in the system, whom it is geared to and what explanations can be generated to meet this need. We then consider when the user should be presented with this information. Last, we consider how objective and subjective measures can be used to evaluate the entire system. This last question is the most encompassing as it will need to evaluate all other issues regarding explainability.
SummaryDesigners of robotic groups are faced with the formidable task of creating effective coord... more SummaryDesigners of robotic groups are faced with the formidable task of creating effective coordination architectures that can deal with collisions due to changing environment conditions and hardware failures. Communication between robots is a mechanism that can at times be helpful in such systems, but can also create a time, energy, or computation overhead that reduces performance. In dealing with this issue, different communication schemes have been proposed ranging from those without any explicit communication, localized algorithms, and centralized or global communicative methods. Finding the optimal communication act is typically an intractable problem in real-world problems. As a result, we argue that at times group designers should use computationally bounded team communication approaches. We propose two such approaches: an algorithm selection approach to communication whereby robots choose between a known group of communication schemes and a parameterized communication frame...
Traditionally, vehicles have been considered as machines that are controlled by humans for the pu... more Traditionally, vehicles have been considered as machines that are controlled by humans for the purpose of transportation. A more modern view is to envision drivers and passengers as actively interacting with a complex automated system. Such interactive activity leads us to consider intelligent and advanced ways of interaction leading to cars that can adapt to their drivers.In this paper, we focus on the Adaptive Cruise Control (ACC) technology that allows a vehicle to automatically adjust its speed to maintain a preset distance from the vehicle in front of it based on the driver’s preferences. Although individual drivers have different driving styles and preferences, current systems do not distinguish among users. We introduce a method to combine machine learning algorithms with demographic information and expert advice into existing automated assistive systems. This method can reduce the interactions between drivers and automated systems by adjusting parameters relevant to the oper...
Robots are being deployed in an increasing variety of environments for longer periods of time. As... more Robots are being deployed in an increasing variety of environments for longer periods of time. As the number of robots grows, they will increasingly need to interact with other robots. Additionally, the number of companies and research laboratories producing these robots is increasing, leading to the situation where these robots may not share a common communication or coordination protocol. While standards for coordination and communication may be created, we expect that robots will need to additionally reason intelligently about their teammates with limited information. This problem motivates the area of ad hoc teamwork in which an agent may potentially cooperate with a variety of teammates in order to achieve a shared goal. This article focuses on a limited version of the ad hoc teamwork problem in which an agent knows the environmental dynamics and has had past experiences with other teammates, though these experiences may not be representative of the current teammates. To tackle this problem, this article introduces a new general-purpose algorithm, PLASTIC, that reuses knowledge learned from previous teammates or provided by experts to quickly adapt to new teammates. This algorithm is instantiated in two forms: 1) PLASTIC-Model-which builds models of previous teammates' behaviors and plans behaviors online using these models and 2) PLASTIC-Policy-which learns policies for cooperating with previous teammates and selects among these policies online. We evaluate PLASTIC on two benchmark tasks: the pursuit domain and robot soccer in the RoboCup 2D simulation domain. Recognizing that a key requirement of ad hoc teamwork is adaptability to previously unseen agents, the tests use more than 40 previously unknown teams on the first task and 7 previously unknown teams on the second. While PLASTIC assumes that there is some degree of similarity between the current and past teammates' behaviors, no steps are taken in the experimental setup to make sure this assumption holds. The teammates were created by a variety of independent developers and were not designed to share any similarities. Nonetheless, the results show that PLASTIC was able to identify and $ This article contains material from 4 prior conference papers [11-14].
In this paper we investigate methods for measuring the expected utility from communicating inform... more In this paper we investigate methods for measuring the expected utility from communicating information in multi-agent planning and scheduling problems. We consider an environment where human teammates can potentially add information to relax constraint information. As these problems are NP-complete, no polynomial algorithms exist for evaluating the impact of either adding or relaxing a certain constraint will have on the global problem. We present a general approach based on a notion we introduce called problem tightness. Distributed agents use this notion to identify those problems which are not overly constrained and, therefore, will not benefit from additional information that would relax those constraints. Finally, agents apply traditional machine learning methods based on their specific local problem attributes to attempt to identify which of the constrained problems will most benefit from added information. We evaluated this approach within a distributed c-TAEMS scheduling domain and found that this approach was effective overall.
An important area of social network research is identifying missing information which is not visi... more An important area of social network research is identifying missing information which is not visible or explicitly represented in the network. Recently, the Missing Node Identification problem was introduced where missing members in the social network structure must be identified. However, previous works did not consider the possibility that information about specific users (nodes) within the network may be known and could be useful in solving this problem. Assuming such information such as user demographic information and users' historical behavior in the network is known, more effective algorithms for the Missing Node Identification problem could potentially be developed. In this paper, we present three algorithms, SAMI-A, SAMI-C and SAMI-N, which leverage this type of information in order to perform significantly better than previous missing node algorithms. However, as each of these algorithms and the parameters within these algorithms often perform better in specific problem instances, a mechanism is needed to select the best algorithm and the best variation within that algorithm. Towards this challenge, we also present OASCA, a novel online selection algorithm. We present results that detail the success of the algorithms presented within this paper.
Autonomous Agents & Multiagent Systems/International Conference on Autonomous Agents, 2004
Abstract: IntroductionTo date, only limited work has been performed on studyinghow performance sc... more Abstract: IntroductionTo date, only limited work has been performed on studyinghow performance scales with the addition of robots togroups. Our work focuses on studying this issue withinone domain, robotic foraging. Foraging has been extensivelystudied, and is formally defined as consisting of locatingtarget items from a search region S, and deliveringthem to a goal region G [3]. Previous work by Rybski
In this paper we investigate methods for analyzing the expected value of adding information in di... more In this paper we investigate methods for analyzing the expected value of adding information in distributed task scheduling problems. As scheduling problems are NP-complete, no polynomial algorithms exist for evaluating the impact a certain constraint, or relaxing the same constraint, will have on the global problem. We present a general approach where local agents can estimate their problem tightness, or how constrained their local subproblem is. This allows these agents to immediately identify many problems which are not constrained, and will not benefit from sending or receiving further information. Next, agents use traditional machine learning methods based on their specific local problem attributes to attempt to identify which of the constrained problems will most benefit from human attention. We evaluated this approach within a distributed cTAEMS scheduling domain and found this approach was overall quite effective.
2009 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technology, 2009
In collaborative systems involving a user and an agent working together on a joint task it may be... more In collaborative systems involving a user and an agent working together on a joint task it may be important to share information in order to determine the appropriate course of action. However, communication between agents and users can create costly user interruptions. One of the most important issue concerning the initiation of information sharing in collaborative systems is the ability to accurately estimate the cost and bene t arising from those interruptions. While cost estimation of interruptions has been previously investigated, these works assumed either a large amount of information was available about each user, or only a small number of states needed consideration. This paper presents a novel synthesis between Collaborative Filtering methods with classi cation algorithms tools to create a fast learning algorithm, MICU. MICU exploits the similarities between users in order to learn from known users to new but similar users and therefore requires less information on each user in compare to other methods. Experimental results indicate the algorithm signi cantly improves system performance even with a small amount of data on each user.
Autonomous agents developed by experts are embedded with the capability to interact well with peo... more Autonomous agents developed by experts are embedded with the capability to interact well with people from different cultures. When designing expert agents intended to interact with autonomous agents developed by Non Game Theory Agents (NGTE), it is beneficial to obtain insights on the behavior of these NGTE agents. Is the behavior of these NGTE agents similar to human behavior from different cultures? This is an important question as such a quality would allow an expert agent interacting with NGTE agents to model them using the same methods that are used to model humans from different cultures. To study this point, we evaluated NGTE agents behavior using a game called the Trust-Revenge game, which is known in social science for capturing different human tendencies. The Trust-Revenge game has a unique subgame-perfect equilibrium strategy profile, however, very rarely do people follow it. We compared the behavior of autonomous agents to the actions of several human demographic groups-one of which is similar to the designers of the autonomous agents. We claim that autonomous agents are similar to human players from various cultures. This enables the use of approaches, developed for handling cultural diversity among humans, to be applied for interaction with NGTE agents. This paper also analyzes additional aspects of autonomous agents behavior and whether composing autonomous agents affects human behavior.
Recent advances in peer to peer (P2P) search algorithms have presented viable structured and unst... more Recent advances in peer to peer (P2P) search algorithms have presented viable structured and unstructured approaches for full-text search. We posit that these existing approaches are each best suited for different types of queries. We present PHIRST, the first system to facilitate effective full-text search within P2P networks. PHIRST works by effectively leveraging between the relative strengths of these approaches. Similar to structured approaches, agents first publish terms within their stored documents. However, frequent terms are quickly identified and not exhaustively stored, resulting in a significantly reduction in the system's storage requirements. During query lookup, agents use unstructured searches to compensate for the lack of fully published terms. Additionally, they explicitly weigh between the costs involved with structured and unstructured approaches, allowing for a significant reduction in query costs. We evaluated the effectiveness of our approach using both real-world and artificial queries. We found that in most situations our approach yields near perfect recall. We discuss the limitations of our system, as well as possible compensatory strategies.
ACM Transactions on Knowledge Discovery from Data, 2014
In recent years, social networks have surged in popularity. One key aspect of social network rese... more In recent years, social networks have surged in popularity. One key aspect of social network research is identifying important missing information that is not explicitly represented in the network, or is not visible to all. To date, this line of research typically focused on finding the connections that are missing between nodes, a challenge typically termed as the link prediction problem . This article introduces the missing node identification problem, where missing members in the social network structure must be identified. In this problem, indications of missing nodes are assumed to exist. Given these indications and a partial network, we must assess which indications originate from the same missing node and determine the full network structure. Toward solving this problem, we present the missing node identification by spectral clustering algorithm (MISC), an approach based on a spectral clustering algorithm, combined with nodes’ pairwise affinity measures that were adopted from...
Agents that interact with humans are known to benefit from integrating behavioral science and exp... more Agents that interact with humans are known to benefit from integrating behavioral science and exploiting the fact that humans are irrational. Therefore, when designing agents for interacting with automated agents, it is crucial to know whether the other agents are acting irrationally and if so to what extent. However, little is known about whether irrationality is found in automated agent design. Do automated agents suffer from irrationality? If so, is it similar in nature and extent to human irrationality? How do agents act in domains where human irrationality is motivated by emotion? This is the first time that extensive experimental evaluation was performed in order to resolve these questions. We evaluated agent rationality (for non-expert agents) in several environments and compared agent actions to human actions. We found that automated agents suffer from the same irrationality that humans display, although to a lesser degree. Automated Agents are integrated into countless environments, such as Electronic commerce, Web crawlers, Military agents, Space Exploration probes and Automated drivers. Due to the high importance of automated multi-agent environments, many competitions were established where automated agents compete with each other in order to achieve a goal . Modeling agents is beneficial for agent-agent interaction [29]. However, building such a model is a complex issue, and furthermore, if the model built is too far from the actual opponent's behavior, using it may become detrimental . How should designers plan their agents when opponent modeling is unavailable? Can any general assumptions be made on automated agents and used for agent design? Research into peoples' behavior has found that people often do not make strictly rational decisions but instead use sub-optimal, bounded policies. This behavior has been attributed to a variety of reasons including: a lack of knowledge of one's own preferences, the effects of the task complexity, framing effects, the interplay between emotion and cognition, the problem of self-control, the value of anticipation, future discounting, anchoring and many other effects . Since people do not usually use fully rational strategies themselves, agents based on game theory approach, which assume rational behavior in humans often perform poorly . Many studies have shown that psychological factors and human decision-making theory are needed in order to develop a good model of true human behavior, which in turn is required for optimizing the performance of agents interacting with humans .
This paper presents a taxonomy of explainability in human-agent systems. We consider fundamental ... more This paper presents a taxonomy of explainability in human-agent systems. We consider fundamental questions about the Why, Who, What, When and How of explainability. First, we define explainability, and its relationship to the related terms of interpretability, transparency, explicitness, and faithfulness. These definitions allow us to answer why explainability is needed in the system, whom it is geared to and what explanations can be generated to meet this need. We then consider when the user should be presented with this information. Last, we consider how objective and subjective measures can be used to evaluate the entire system. This last question is the most encompassing as it will need to evaluate all other issues regarding explainability.
This paper presents objective metrics for how explainable artificial intelligence (XAI) can be qu... more This paper presents objective metrics for how explainable artificial intelligence (XAI) can be quantified. Through an overview of current trends, we show that many explanations are generated post-hoc and independent of the agent's logical process, which in turn creates explanations with limited meaning as they lack transparency and fidelity. While user studies are a known basis for evaluating XAI, studies that do not consider objective metrics for evaluating XAI may have limited meaning and may suffer from confirmation bias, particularly if they use low fidelity explanations unnecessarily. To avoid this issue, this paper suggests a paradigm shift in evaluating XAI that focuses on metrics that quantify the explanation itself and its appropriateness given the XAI goal. We suggest four such metrics based on performance differences, D, between the explanation's logic and the agent's actual performance, the number of rules, R, outputted by the explanation, the number of features, F , used to generate that explanation, and the stability, S, of the explanation. We believe that user studies that focus on these metrics in their evaluations are inherently more valid and should be integrated in future XAI research.
Adaptive Agents and Multi-Agents Systems, May 2, 2011
Creating agents that properly simulate and interact with people is critical for many applications... more Creating agents that properly simulate and interact with people is critical for many applications. Towards creating these agents, models are needed that quickly and accurately predict how people behave in a variety of domains and problems. This paper explores how one bounded rationality theory, Aspiration Adaptation Theory (AAT), can be used to aid in this task. We extensively studied two types of problems -a relatively simple optimization problem and two complex negotiation problems. We compared the predictive capabilities of traditional learning methods with those where we added key elements of AAT and other optimal and bounded rationality models. Within the extensive empirical studies we conducted, we found that machine learning models combined with AAT were most effective in quickly and accurately predicting people's behavior.
Adaptive Agents and Multi-Agents Systems, May 5, 2014
Agents that interact with humans are known to benefit from modeling them. Therefore, when designi... more Agents that interact with humans are known to benefit from modeling them. Therefore, when designing agents intended for interaction with automated agents, it is crucial to model the other agents. However, little is known about how to model automated agents and in particular non-expert agents. Are automated agents to be modeled the same way that an agent models humans? Or does a separate model for interacting with automated agents need to be developed? We evaluate automated agent behavior (for non-expert agents) using a game called the Trust-Revenge game, which is known in social science for capturing different human tendencies. The Trust-Revenge game has a unique sub game-perfect equilibrium, however, very rarely do people follow it. We compared the behavior of automated agents to that of human actions in several demographic groups (including a group which is similar but not identical to the designers of the automated agents). We show that differences between automated agents' and humans' behavior are similar to differences between different human cultures.
Internet social networks have become a ubiquitous application allowing people to easily share tex... more Internet social networks have become a ubiquitous application allowing people to easily share text, pictures, and audio and video files. Popular networks include WhatsApp, Facebook, Reddit and LinkedIn. We present an extensive study of the usage of the WhatsApp social network, an Internet messaging application that is quickly replacing SMS messaging. In order to better understand people's use of the network, we provide an analysis of over 6 million messages from over 100 users, with the objective of building demographic prediction models using activity data. We performed extensive statistical and numerical analysis of the data and found significant differences in WhatsApp usage across people of different genders and ages. We also inputted the data into the Weka data mining package and studied models created from decision tree and Bayesian network algorithms. We found that different genders and age demographics had significantly different usage habits in almost all message and gr...
This paper presents an extensive study of the usage of the WhatsApp social network, an Internet m... more This paper presents an extensive study of the usage of the WhatsApp social network, an Internet messaging application that is quickly replacing SMS messaging. It is based on the analysis of over 4 million messages from nearly 100 users that we collected in order to understand people’s use of the network. We believe that this is the first in-depth study of the properties of WhatsApp messages with an emphasis of noting differences across different age and gender demographic groups. It is also the first to use statistical and data analytic tools in this analysis. We found that different genders and age demographics had significantly different usage habits in almost all message and group attributes. These differences facilitate the development of user prediction models based on data mining tools. We illustrate this by developing several prediction models such as for a person’s gender or approximate age. We also noted differences in users’ group behavior. We created group behaviorial mod...
BACKGROUND Social networks on the Internet have become ubiquitous applications that allow people ... more BACKGROUND Social networks on the Internet have become ubiquitous applications that allow people to easily share text, pictures, and audio and video files. Popular networks include WhatsApp, Facebook, Reddit, and LinkedIn.
This paper presents a taxonomy of explainability in Human-Agent Systems. We consider fundamental ... more This paper presents a taxonomy of explainability in Human-Agent Systems. We consider fundamental questions about the Why, Who, What, When and How of explainability. First, we define explainability, and its relationship to the related terms of interpretability, transparency, explicitness, and faithfulness. These definitions allow us to answer why explainability is needed in the system, whom it is geared to and what explanations can be generated to meet this need. We then consider when the user should be presented with this information. Last, we consider how objective and subjective measures can be used to evaluate the entire system. This last question is the most encompassing as it will need to evaluate all other issues regarding explainability.
SummaryDesigners of robotic groups are faced with the formidable task of creating effective coord... more SummaryDesigners of robotic groups are faced with the formidable task of creating effective coordination architectures that can deal with collisions due to changing environment conditions and hardware failures. Communication between robots is a mechanism that can at times be helpful in such systems, but can also create a time, energy, or computation overhead that reduces performance. In dealing with this issue, different communication schemes have been proposed ranging from those without any explicit communication, localized algorithms, and centralized or global communicative methods. Finding the optimal communication act is typically an intractable problem in real-world problems. As a result, we argue that at times group designers should use computationally bounded team communication approaches. We propose two such approaches: an algorithm selection approach to communication whereby robots choose between a known group of communication schemes and a parameterized communication frame...
Traditionally, vehicles have been considered as machines that are controlled by humans for the pu... more Traditionally, vehicles have been considered as machines that are controlled by humans for the purpose of transportation. A more modern view is to envision drivers and passengers as actively interacting with a complex automated system. Such interactive activity leads us to consider intelligent and advanced ways of interaction leading to cars that can adapt to their drivers.In this paper, we focus on the Adaptive Cruise Control (ACC) technology that allows a vehicle to automatically adjust its speed to maintain a preset distance from the vehicle in front of it based on the driver’s preferences. Although individual drivers have different driving styles and preferences, current systems do not distinguish among users. We introduce a method to combine machine learning algorithms with demographic information and expert advice into existing automated assistive systems. This method can reduce the interactions between drivers and automated systems by adjusting parameters relevant to the oper...
Robots are being deployed in an increasing variety of environments for longer periods of time. As... more Robots are being deployed in an increasing variety of environments for longer periods of time. As the number of robots grows, they will increasingly need to interact with other robots. Additionally, the number of companies and research laboratories producing these robots is increasing, leading to the situation where these robots may not share a common communication or coordination protocol. While standards for coordination and communication may be created, we expect that robots will need to additionally reason intelligently about their teammates with limited information. This problem motivates the area of ad hoc teamwork in which an agent may potentially cooperate with a variety of teammates in order to achieve a shared goal. This article focuses on a limited version of the ad hoc teamwork problem in which an agent knows the environmental dynamics and has had past experiences with other teammates, though these experiences may not be representative of the current teammates. To tackle this problem, this article introduces a new general-purpose algorithm, PLASTIC, that reuses knowledge learned from previous teammates or provided by experts to quickly adapt to new teammates. This algorithm is instantiated in two forms: 1) PLASTIC-Model-which builds models of previous teammates' behaviors and plans behaviors online using these models and 2) PLASTIC-Policy-which learns policies for cooperating with previous teammates and selects among these policies online. We evaluate PLASTIC on two benchmark tasks: the pursuit domain and robot soccer in the RoboCup 2D simulation domain. Recognizing that a key requirement of ad hoc teamwork is adaptability to previously unseen agents, the tests use more than 40 previously unknown teams on the first task and 7 previously unknown teams on the second. While PLASTIC assumes that there is some degree of similarity between the current and past teammates' behaviors, no steps are taken in the experimental setup to make sure this assumption holds. The teammates were created by a variety of independent developers and were not designed to share any similarities. Nonetheless, the results show that PLASTIC was able to identify and $ This article contains material from 4 prior conference papers [11-14].
In this paper we investigate methods for measuring the expected utility from communicating inform... more In this paper we investigate methods for measuring the expected utility from communicating information in multi-agent planning and scheduling problems. We consider an environment where human teammates can potentially add information to relax constraint information. As these problems are NP-complete, no polynomial algorithms exist for evaluating the impact of either adding or relaxing a certain constraint will have on the global problem. We present a general approach based on a notion we introduce called problem tightness. Distributed agents use this notion to identify those problems which are not overly constrained and, therefore, will not benefit from additional information that would relax those constraints. Finally, agents apply traditional machine learning methods based on their specific local problem attributes to attempt to identify which of the constrained problems will most benefit from added information. We evaluated this approach within a distributed c-TAEMS scheduling domain and found that this approach was effective overall.
An important area of social network research is identifying missing information which is not visi... more An important area of social network research is identifying missing information which is not visible or explicitly represented in the network. Recently, the Missing Node Identification problem was introduced where missing members in the social network structure must be identified. However, previous works did not consider the possibility that information about specific users (nodes) within the network may be known and could be useful in solving this problem. Assuming such information such as user demographic information and users' historical behavior in the network is known, more effective algorithms for the Missing Node Identification problem could potentially be developed. In this paper, we present three algorithms, SAMI-A, SAMI-C and SAMI-N, which leverage this type of information in order to perform significantly better than previous missing node algorithms. However, as each of these algorithms and the parameters within these algorithms often perform better in specific problem instances, a mechanism is needed to select the best algorithm and the best variation within that algorithm. Towards this challenge, we also present OASCA, a novel online selection algorithm. We present results that detail the success of the algorithms presented within this paper.
Autonomous Agents & Multiagent Systems/International Conference on Autonomous Agents, 2004
Abstract: IntroductionTo date, only limited work has been performed on studyinghow performance sc... more Abstract: IntroductionTo date, only limited work has been performed on studyinghow performance scales with the addition of robots togroups. Our work focuses on studying this issue withinone domain, robotic foraging. Foraging has been extensivelystudied, and is formally defined as consisting of locatingtarget items from a search region S, and deliveringthem to a goal region G [3]. Previous work by Rybski
In this paper we investigate methods for analyzing the expected value of adding information in di... more In this paper we investigate methods for analyzing the expected value of adding information in distributed task scheduling problems. As scheduling problems are NP-complete, no polynomial algorithms exist for evaluating the impact a certain constraint, or relaxing the same constraint, will have on the global problem. We present a general approach where local agents can estimate their problem tightness, or how constrained their local subproblem is. This allows these agents to immediately identify many problems which are not constrained, and will not benefit from sending or receiving further information. Next, agents use traditional machine learning methods based on their specific local problem attributes to attempt to identify which of the constrained problems will most benefit from human attention. We evaluated this approach within a distributed cTAEMS scheduling domain and found this approach was overall quite effective.
2009 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technology, 2009
In collaborative systems involving a user and an agent working together on a joint task it may be... more In collaborative systems involving a user and an agent working together on a joint task it may be important to share information in order to determine the appropriate course of action. However, communication between agents and users can create costly user interruptions. One of the most important issue concerning the initiation of information sharing in collaborative systems is the ability to accurately estimate the cost and bene t arising from those interruptions. While cost estimation of interruptions has been previously investigated, these works assumed either a large amount of information was available about each user, or only a small number of states needed consideration. This paper presents a novel synthesis between Collaborative Filtering methods with classi cation algorithms tools to create a fast learning algorithm, MICU. MICU exploits the similarities between users in order to learn from known users to new but similar users and therefore requires less information on each user in compare to other methods. Experimental results indicate the algorithm signi cantly improves system performance even with a small amount of data on each user.
Autonomous agents developed by experts are embedded with the capability to interact well with peo... more Autonomous agents developed by experts are embedded with the capability to interact well with people from different cultures. When designing expert agents intended to interact with autonomous agents developed by Non Game Theory Agents (NGTE), it is beneficial to obtain insights on the behavior of these NGTE agents. Is the behavior of these NGTE agents similar to human behavior from different cultures? This is an important question as such a quality would allow an expert agent interacting with NGTE agents to model them using the same methods that are used to model humans from different cultures. To study this point, we evaluated NGTE agents behavior using a game called the Trust-Revenge game, which is known in social science for capturing different human tendencies. The Trust-Revenge game has a unique subgame-perfect equilibrium strategy profile, however, very rarely do people follow it. We compared the behavior of autonomous agents to the actions of several human demographic groups-one of which is similar to the designers of the autonomous agents. We claim that autonomous agents are similar to human players from various cultures. This enables the use of approaches, developed for handling cultural diversity among humans, to be applied for interaction with NGTE agents. This paper also analyzes additional aspects of autonomous agents behavior and whether composing autonomous agents affects human behavior.
Recent advances in peer to peer (P2P) search algorithms have presented viable structured and unst... more Recent advances in peer to peer (P2P) search algorithms have presented viable structured and unstructured approaches for full-text search. We posit that these existing approaches are each best suited for different types of queries. We present PHIRST, the first system to facilitate effective full-text search within P2P networks. PHIRST works by effectively leveraging between the relative strengths of these approaches. Similar to structured approaches, agents first publish terms within their stored documents. However, frequent terms are quickly identified and not exhaustively stored, resulting in a significantly reduction in the system's storage requirements. During query lookup, agents use unstructured searches to compensate for the lack of fully published terms. Additionally, they explicitly weigh between the costs involved with structured and unstructured approaches, allowing for a significant reduction in query costs. We evaluated the effectiveness of our approach using both real-world and artificial queries. We found that in most situations our approach yields near perfect recall. We discuss the limitations of our system, as well as possible compensatory strategies.
ACM Transactions on Knowledge Discovery from Data, 2014
In recent years, social networks have surged in popularity. One key aspect of social network rese... more In recent years, social networks have surged in popularity. One key aspect of social network research is identifying important missing information that is not explicitly represented in the network, or is not visible to all. To date, this line of research typically focused on finding the connections that are missing between nodes, a challenge typically termed as the link prediction problem . This article introduces the missing node identification problem, where missing members in the social network structure must be identified. In this problem, indications of missing nodes are assumed to exist. Given these indications and a partial network, we must assess which indications originate from the same missing node and determine the full network structure. Toward solving this problem, we present the missing node identification by spectral clustering algorithm (MISC), an approach based on a spectral clustering algorithm, combined with nodes’ pairwise affinity measures that were adopted from...
Agents that interact with humans are known to benefit from integrating behavioral science and exp... more Agents that interact with humans are known to benefit from integrating behavioral science and exploiting the fact that humans are irrational. Therefore, when designing agents for interacting with automated agents, it is crucial to know whether the other agents are acting irrationally and if so to what extent. However, little is known about whether irrationality is found in automated agent design. Do automated agents suffer from irrationality? If so, is it similar in nature and extent to human irrationality? How do agents act in domains where human irrationality is motivated by emotion? This is the first time that extensive experimental evaluation was performed in order to resolve these questions. We evaluated agent rationality (for non-expert agents) in several environments and compared agent actions to human actions. We found that automated agents suffer from the same irrationality that humans display, although to a lesser degree. Automated Agents are integrated into countless environments, such as Electronic commerce, Web crawlers, Military agents, Space Exploration probes and Automated drivers. Due to the high importance of automated multi-agent environments, many competitions were established where automated agents compete with each other in order to achieve a goal . Modeling agents is beneficial for agent-agent interaction [29]. However, building such a model is a complex issue, and furthermore, if the model built is too far from the actual opponent's behavior, using it may become detrimental . How should designers plan their agents when opponent modeling is unavailable? Can any general assumptions be made on automated agents and used for agent design? Research into peoples' behavior has found that people often do not make strictly rational decisions but instead use sub-optimal, bounded policies. This behavior has been attributed to a variety of reasons including: a lack of knowledge of one's own preferences, the effects of the task complexity, framing effects, the interplay between emotion and cognition, the problem of self-control, the value of anticipation, future discounting, anchoring and many other effects . Since people do not usually use fully rational strategies themselves, agents based on game theory approach, which assume rational behavior in humans often perform poorly . Many studies have shown that psychological factors and human decision-making theory are needed in order to develop a good model of true human behavior, which in turn is required for optimizing the performance of agents interacting with humans .
Uploads
Papers by Avi Rosenfeld