Agents and Artificial Intelligence

This paper reports how our perception of virtual agents differ by the combination of the gender of their appearances and gestures. We examined how we perceive masculinity and femininity of agents and how our perception of agent’s gender affect our impression of the agent. Human-human interactions among Japanese undergraduate students were analyzed with respect to usage of gender-specific self-adaptors in a pre-experiment. Based on the results, a male and a female agent were animated to show these extracted self-adaptors. Evaluation of the interactions with the agents that exhibit self-adaptors typically exhibited by Japanese human male and female indicated that there are cross gender interactions between participants’ gender and agents’ gender. Male participants showed more favorable impressions on agents that display feminine self-adaptors than masculine ones performed by the female agent, while female participants showed rigorous impressions toward feminine self-adaptors. Although the obtained results were limited to one culture and narrow age range, these results implies there is a possibility that the combination of male appearance and masculine gestures is “safer” in order to facilitate neutral impressions and avoid any cross gender interactions made by the gender of human users. Designers of virtual agents should consider gender of appearance and gesture animations of virtual agents, and make them customizable according to the user’s gender and preferences.

In this paper we introduce a novel application of model checking to find optimal planning solutions for a flow production system. Originally controlled by a multiagent system, the production system consists of autonomous products and asynchronous production stations with limited space for waiting products. In this work, we present two different approaches of application of the Spin model checker to optimize throughput in the given production system. Instead of mapping the multiagent system directly, we model the production line itself as a set of communicating processes. Each communication channel between two processes represents a one-way monorail connection from one station to another. Experiments show that both approaches derive valid and optimized plans with several thousands of steps using constrained branch-and-bound. However, experiments also indicate individual advantages of both approaches.

This paper proposes a control method for in agents by switching their behavioral strategy between rationality and reciprocity depending on their internal states to achieve efficient team formation. Advances in computer science, telecommunications, and electronic devices have led to proposals of a variety of services on the Internet that are achieved by teams of different agents. To provide these services efficiently, the tasks to achieve them must be allocated to appropriate agents that have the required capabilities, and the agents must not be overloaded. Furthermore, agents have to adapt to dynamic environments, especially to frequent changes in workload. Conventional decentralized allocation methods often lead to conflicts in large and busy environments because high-capability agents are likely to be identified as the best team member by many agents, resulting in the entire system becoming inefficient due to the concentration of task allocation when the workload becomes high. Our proposed agents switch their strategies in accordance with their local evaluation to avoid conflicts occurring in busy environments. They also establish an organization in which a number of groups are autonomously generated in a bottom-up manner on the basis of dependability to avoid conflicts in advance while ignoring tasks allocated by undependable/unreliable agents. We experimentally evaluated our method in static and dynamic environments where the number of tasks varied.

Our aim is to understand reviews from the point of view of the arguments they contain, and then do a first step from how arguments are distributed in such reviews towards the behaviour of the reviewers that posted them. We consider 253 reviews of a selected product (a ballet tutu for kids), extracted from the “Clothing, Shoes and Jeweller” section of Amazon.​com. We explode these reviews into arguments, and we study how their characteristics, e.g., the distribution of positive (in favour of purchase) and negative ones (against purchase), change through a period of four years. Among other results, we discover that negative arguments tend to permeate also positive reviews. As a second step, by using such observations and distributions, we successfully replicate the reviewers’ behaviour by simulating the review-posting process from their basic components, i.e., the arguments themselves.

Natural language statements uttered in diagnosis, but more general in daily life are usually graded, i.e., are associated with a degree of uncertainty about the validity of an assessment and is often expressed through specific words in natural language. In this paper, we look into a representation of such graded statements by presenting a simple non-standard modal logic which comes with a set of modal operators, directly associated with the words indicating the uncertainty and interpreted through confidence intervals in the model theory. We complement the model theory by a set of RDFS-/OWL 2 RL-like entailment (if-then) rules, acting on the syntactic representation of modalized statements. After that, we extend the modal statements by transaction time, in order to implement a notion of temporal change. Our interest in such a formalization is related to the use of OWL as the de facto language in today’s ontologies and its weakness to represent and reason about assertional knowledge that is uncertain and that changes over time.

This paper proposes a design framework for a personalized multi-agent recommender system. More precisely, the proposed framework is a multi-context based recommender system that takes into account user preferences to generate a plan satisfying those preferences. Agents in this framework have a Belief-Desire-Intention (BDI) component based on the well-known BDI architecture. These BDI agents are empowered with cognitive capabilities in order to interact with others agents. They are also able to adapt to the environment changes and to the information coming from other agents. The architecture includes also a planning module based on ontologies in order to represent and reason about plans and intentions. The applicability of the proposed model is shown through a simulation in the NetLogo environment.

Predictive scheduling is a frequently executed task within the control process of energy grids. Relying on different predictions, planning results are naturally subject to uncertainties. Robust proactive planning of day-ahead real power provision must incorporate uncertainty in feasibility when trading off different schedules against each other during the predictive planning phase. Deviations from the expected initial operational state of an energy unit may easily foil a planned schedule commitment and provoke the need for costly ancillary services. The integration of confidence information into the optimization model allows for a consideration of uncertainty at planning time; resulting in more robust plans. Hence, control power and costs arising from deviations from agreed energy product delivery can be minimized. Integrating uncertainty information can be easily done when using a surrogate model. We extend an existing surrogate model that has been successfully used in energy management for checking feasibility during constraint-based optimization. The surrogate is extended to incorporate confidence scores based on expected feasibility under changed operational conditions. We compare the new surrogate model with the old one and demonstrate the superiority of the new model by results from several simulation studies.

Argumentation mining aims at automatically extracting natural language arguments from textual documents. In the last years, it has become a hot topic due to its potential in processing information originating from the Web in innovative ways. In this paper, we propose to apply the argument mining pipeline to the text exploration task. First, starting from the arguments put forward in online debates, we introduce bipolar entailment graphs to predict the relation among the textual arguments, i.e., entailment or non entailment relation. Second, we exploit the well know formalism called abstract dialectical frameworks to define acceptance conditions answering the needs of the text exploration task. The evaluation of the proposed approach shows its feasibility.

Classification of high-dimensional time series with imbalanced classes is a challenging task. For such classification tasks, the cascade classifier has been proposed. The cascade classifier tackles high-dimensionality and imbalance by splitting the classification task into several low-dimensional classification tasks and aggregating the intermediate results. Therefore the high-dimensional data set is projected onto low-dimensional subsets. But these subsets can employ unfavorable and not representative data distributions, that hamper classifiction again. Data preprocessing can overcome these problems. Small improvements in the low-dimensional data subsets of the cascade classifier lead to an improvement of the aggregated overall results. We present two data preprocessing methods, instance selection and outlier generation. Both methods are based on point distances in low-dimensional space. The instance selection method selects representative feasible examples and the outlier generation method generates artificial infeasible examples near the class boundary. In an experimental study, we analyse the precision improvement of the cascade classifier due to the presented data preprocessing methods for power production time series of a micro Combined Heat and Power plant and an artificial and complex data set. The precision increase is due to an increased selectivity of the learned decision boundaries. This paper is an extended version of [19], where we have proposed the two data preprocessing methods. In this paper we extend the analysis of both algorithms by a parameter sensitivity analysis of the distance parameters from the preprocessing methods. Both distance parameters depend on each other and have to be chosen carefully. We study the influence of these distance parameters on the classification precision of the cascade model and derive parameter fitting rules for the \(\mu \)CHP data set. The experiments yield a region of optimal parameter value combinations leading to a high classification precision.

Min-based possibilistic influence diagrams offer a compact modeling of decision problems under uncertainty. Uncertainty and preferential relations are expressed on the same structure by using ordinal data. In many applications, it may be natural to represent expert knowledge and preferences separately and treat all nodes similarly. This work shows how an influence diagram can be equivalently represented by two possibilistic networks: the first one represents knowledge of an agent and the second one represents agent’s preferences. Thus, the decision evaluation process is based on more compact possibilistic network. Then, we show that the computation of sequential optimal decisions (strategy) comes down to compute a normalization degree of the junction tree associated with the graph representing the fusion of agents beliefs and its preferences resulting from the proposed decomposition process.

Recent advancements in computing technology allowed both scientific and business applications to produce large datasets with increasing complexity and dimensionality. Clustering algorithms are useful in analyzing these large datasets but often fall short to provide completely satisfactory results. Integrating clustering and visualization not only yields better clustering results but also leads to a higher degree of confidence in the findings. Self-Organizing Map (SOM) is a neural network model which is used to obtain a topology-preserving mapping from the (usually high dimensional) input/feature space to an output/map space of fewer dimensions (usually two or three in order to facilitate visualization). Neurons in the output space are connected with each other but this structure remains fixed throughout training and learning is achieved through the updating of neuron reference vectors in feature space. Despite the fact that growing variants of SOM overcome the fixed structure limitation, they increase computational cost and also do not allow the removal of a neuron after its introduction. In this paper, a variant of SOM is presented called AMSOM (Adaptive Moving Self-Organizing Map) that on the one hand creates a more flexible structure where neuron positions are dynamically altered during training and on the other hand tackles the drawback of having a predefined grid by allowing neuron addition and/or removal during training. Experimental evaluation on different literature datasets with diverse characteristics improves SOM training performance, leads to a better visualization of the input dataset, and provides a framework for determining the optimal number and structure of neurons as well as the optimal number of clusters.

Multi-agent Plan Recognition (MPAR) infers teams and their goals from observed actions of individual agents. The complexity of creating a priori plan libraries significantly increases to account for diversity of action sequences different team structures may exhibit. A key challenge in MPAR is effectively pruning the joint search space of agent to team compositions and goal to team assignments. Here, we describe discrete Multi-agent Plan Recognition as Planning (MAPRAP), which extends Ramirez and Geffner’s Plan Recognition as Planning (PRAP) approach to multi-agent domains. Instead of a plan library, MAPRAP uses the planning domain and synthesizes plans to achieve hypothesized goals with additional constraints for suspected team composition and previous observations. By comparing costs of plans, MAPRAP identifies feasible interpretations that explain the teams and plans observed. We establish a performance profile for discrete MAPRAP in a multi-agent blocks-world domain. We evaluated precision, accuracy, and recall after each observation. We compare two pruning strategies to dampen the explosion of hypotheses tested. Aggressive pruning averages 1.05 plans synthesized per goal per time step for multi-agent scenarios vice 0.56 for single agent scenarios.

In this paper we study Secrecy-Preserving Query Answering problem under Open World Assumption (OWA) for \(\mathcal {EL}^+\) Knowledge Bases (KBs). First we compute some consequences of ABox (\(\mathcal {A}\)) and TBox (\(\mathcal {T}\)) denoted by \(\mathcal {A}^*\) and \(\mathcal {T}^*\) respectively. A secrecy set of a querying agent is subset \(\mathbb {S}\) of \(\mathcal {A}^*\cup \mathcal {T}^*\) which the agent is not allowed to access. Next we compute envelopes which provide logical protection to the secrecy set against the reasoning of the querying agent. Once envelopes are computed, they are used to efficiently answer assertional and GCI queries without compromising the secret information in \(\mathbb {S}\). When the querying agent asks a query q, the reasoner answers “Yes” if KB \(\models q\) and q does not belong to the envelopes; otherwise, the reasoner answers “Unknown”. Being able to answer “Unknown” plays a key role in protecting secrecy under OWA. Since we are not computing all the consequences of the KB, answers to the queries based on just \(\mathcal {A}^*\) and \(\mathcal {T}^*\) could be erroneous. To fix this problem, we further augment our algorithms to make the query answering procedure foolproof.

Eliciting representative membership functions is one of the fundamental steps in applications of fuzzy theory. This paper investigates an unsupervised approach that incorporates variable bandwidth mean-shift and robust statistics for generating fuzzy membership functions. The approach automatically learns the number of representative functions from the underlying data distribution. Given a specific membership function, the approach then works out the associated parameters of the specific membership function. Our evaluation of the proposed approach consists of comparisons with two other techniques in terms of (i) parameterising MFs for attributes with different distributions, and (ii) classification performance of a fuzzy rule set that was developed using the parameterised output of these techniques. This evaluation involved its application using the trapezoidal and the triangular membership functions. Results demonstrate that the generated membership functions can better separate the underlying distributions and classifiers constructed using the proposed method of generating membership function outperformed three other classifiers that used different approaches for parameterisation of the attributes.

This paper considers the problem of facilitating coalition formation in self-interested multi-agent environments. To successfully form a coalition, agents must collectively agree on the monetary amount to charge for completion of a task as well as the distribution of subtasks within the coalition. The problem is accentuated as different subtasks have various degrees of difficulty and the agents do not possess perfect information. That is, an agent is uncertain of the true monetary requirement of other agents for completing subtasks. These complexities, coupled with the self-interested nature of agents, can inhibit or even prevent the formation of coalitions in such a real-world setting. As a solution we present an auction-based protocol called ACCORD. ACCORD facilitates coalition formation by promoting the adoption of cooperative behaviour amongst agents as a means of overcoming the complexities outlined above. Through extensive empirical analysis we analyse two variations of the ACCORD protocol and demonstrate that cooperative and fair behaviour is dominant and any agents deviating from this behaviour suffer a degradation in performance.

It is known that the comprehension of spatial prepositions involves the deployment of visual attention. For example, consider the sentence “The salt is to the left of the stove”. Researchers [29, 30] have theorized that people must shift their attention from the stove (the reference object, RO) to the salt (the located object, LO) in order to comprehend the sentence. Such a shift was also implicitly assumed in the Attentional Vector Sum (AVS) model by [35], a cognitive model that computes an acceptability rating for a spatial preposition given a display that contains an RO and an LO. However, recent empirical findings showed that a shift from the RO to the LO is not necessary to understand a spatial preposition ([3], see also [15, 38]). In contrast, these findings suggest that people perform a shift in the reverse direction (i.e., from the LO to the RO). Thus, we propose the reversed AVS (rAVS) model, a modified version of the AVS model in which attention shifts from the LO to the RO. We assessed the AVS and the rAVS model on the data from [35] using three model simulation methods. Our simulations show that the rAVS model performs as well as the AVS model on these data while it also integrates the recent empirical findings. Moreover, the rAVS model achieves its good performance while being less flexible than the AVS model. (This article is an updated and extended version of the paper [23] presented at the 8th International Conference on Agents and Artificial Intelligence in Rome, Italy. The authors would like to thank Holger Schultheis for helpful discussions about the additional model simulation.)

Detecting visible as well as invisible objects of interest in real-world scenes is crucial in new-generation video-surveillance. For this purpose, we design a fully intelligent system incorporating semantic, symbolic, and grounded information. In particular, we conceptualize temporal representations we use together with spatial and visual information in our multi-view tracking system. It uses them for automated reasoning and induction of knowledge about the multiple views of the studied scene, in order to automatically detect salient or hidden objects of interest. Tests on standard datasets demonstrated the efficiency and accuracy of our proposed approach.