EAA 2024 Abstracts

Area 1 - EAA
Full Papers Short Papers

Area 1 - EAA

Full Papers
Paper Nr: 5
Title:

Distributed Theory of Mind in Multi-Agent Systems
Authors:

Heitor Henrique da Silva, Michele Rocha, Guilherme Trajano, Analúcia Schiaffino Morales, Stefan Sarkadi and Alison R. Panisson
Abstract: Theory of Mind is a concept from developmental psychology which elucidates how humans mentalise. More specifically, it describes how humans ascribe mental attitudes to others and how they reason about these mental attitudes. In the area of Artificial Intelligence, Theory of Mind serves as a fundamental pillar in the design of intelligent artificial agents that are supposed to coexist with humans within a hybrid society. Having the ability to mentalise, these artificial agents could potentially exhibit a range of advanced capabilities that underlie meaningful communication, including empathy and the capacity to better understanding the meaning behind the utterances others make. In this paper, we propose a distributed theory of mind approach in multi-agent systems, in which agents and human users share evidence to reach more supported conclusions about each other’s mental attitudes. We demonstrate our approach in a scenario of stress detection, in which personal agents infer whether their users are stressed or not according to the distributed theory of mind approach.
Download
Paper Nr: 6
Title:

A Proposal for Selecting the Most Value-Aligned Preferences in Decision-Making Using Agreement Solutions
Authors:

Aarón López-García
Abstract: Decision-making is mostly subjected to conflict of interest. To solve such a concern, we propose a methodology to generate agreement solutions that determine the most value-aligned preference system according to the stakeholders. These preferences are represented as a weighting scheme that produces a ranking system through the TOPSIS technique. Such an agreement is obtained utilizing an unweighted multicriteria strategy and the least-squares approximation. As a result, this weighting vector is an objective data-driven solution, thus giving empirical evidence and adaptability learning in our proposal. The given solution is also explainable and scalable per se thanks to the multicriteria technique selected. The agreement weight is used to perform a ranking system that solves the decision problem considering the value preferences of the stakeholders. We performed an illustrative example to show the different steps from which the decision problem must be posed to be resolved. We conclude that our proposal is quite effective for solving value-based decision problems in which conflicts of interest arise among affective agents. Moreover, we show the interpretation of the agreement solution and its use in decision-making.
Download
Paper Nr: 7
Title:

Exploring the Relationship Between Emotions and Norms in Decision-Making Processes of Intelligent Agents
Authors:

Joaquin Taverner, Carmengelys Cordova, Elena D. Val, Soledad Valero and Estefania Argente
Abstract: In this paper, we explore the relationship between norms and emotions, examining the fundamental implications that they entail in the development of future models of reasoning and decision-making for intelligent agents based on BDI. Our approach focuses on assessing the impact of anticipated emotions, self-image, and social-image projection as well as utility factors on the complex decision-making processes that an agent faces when deciding whether to comply with or violate a norm. To this end, we propose the use of two types of anticipated emotions, self-conscious, which shape the personal self-image, and social emotions, which are displayed by other agents in the environment and are used to estimate the social image. To represent the agent’s emotional state, we present a new model based on the pleasure dimension in which we represent the self-conscious emotions of Pride and Guilt. Using the language for intelligent agents AgentSpeak, we propose a syntax for defining the norms in the agent’s code. We show a new reasoning cycle based on the BDI model in which we add new functionalities to accommodate affective and normative processes. Affective processes support modifying the agent’s emotional state as well as estimating anticipated emotions and computing self-image and social image. Normative processes allow the instantiation of active norms and normative reasoning.
Download
Paper Nr: 8
Title:

Towards Efficient Driver Distraction Detection with DARTS-Optimized Lightweight Models
Authors:

Yassamine Lala Bouali, Olfa Ben Ahmed, Smaine Mazouzi and Abbas Bradai
Abstract: Driver Distraction is, increasingly, one of the major causes of road accidents. Distractions can be caused by activities that may shift the driver’s attention and potentially evoke negative emotional states. Recently, there has been notable interest in Driver Assistance Systems (DAS) designed for Driver Distraction Detection (DDD). These systems focus on improving both safety and driver comfort by issuing alerts for potential hazards. Recent advancements in DAS have prominently incorporated deep learning techniques, showcasing a shift towards sophisticated and intelligent approaches for enhanced performance and functionality. However, model architecture design is mainly based on expert knowledge and empirical evaluations, which are time-consuming and resource-intensive. Hence, it is hard to design a model that is both efficient and accurate at the same time. This paper presents a Neural Architecture Search (NAS)-based approach for efficient deep CNN design for DDD. The proposed approach leverages RGB images to train a lightweight model with few parameters and high recognition accuracy. Experimental validation is performed on two driver distraction benchmark datasets, demonstrating that the proposed model outperforms state-of-the-art models in terms of efficiency while maintaining competitive accuracy. We report 99.08% and 93.23% with model parameter numbers equal to 0.10 and 0.14 Million parameters for respectively SFD and AUC datasets. The obtained architectures are both accurate and lightweight for DDD.
Download
Short Papers
Paper Nr: 9
Title:

Exploring Text-Generating Large Language Models (LLMs) for Emotion Recognition in Affective Intelligent Agents
Authors:

Aaron Pico, Emilio Vivancos, Ana Garcia-Fornes and Vicente Botti
Abstract: An intelligent agent interacting with a individual will be able to improve its communication with its interlocutor if the agent adapts its behavior according to the individual’s emotional state. In order to do this, it is necessary for the agent to be able to detect the individual’s emotional state through the content of the conversation the agent has with the individual. This paper investigates the application of text-generating Large Language Models (LLMs) for emotion recognition in dialogue settings with the aim of generating emotional knowledge, in the form of beliefs, that can be used by a BDI emotional agent. We compare the performance of several LLMs in recognizing the emotions that an affective BDI agent can employ in its reasoning. Results demonstrate the promising capabilities of diverse models in a Zero-shot prediction (without training and without examples), showcasing the potential for LLMs in emotion recognition tasks. The study advocates for further refinement of LLMs to balance accuracy and efficiency, paving the way for their integration into diverse intelligent agent applications.
Download
Paper Nr: 10
Title:

An Interpretable Machine Learning Approach for Identifying Occupational Stress in Healthcare Professionals
Authors:

Milena S. Fernandes, Roberto R. Filho, Iwens Sene-Junior, Stefan Sarkadi, Alison R. Panisson and Analúcia S. Morales
Abstract: In the last few years, several scientific studies have shown that occupational stress has a significant impact on workers, particularly those in the healthcare sector. This stress is caused by an imbalance between work conditions, the worker’s ability to perform their tasks, and the social support they receive from colleagues and management professionals. Researchers have explored occupational stress as part of a broader study on affective systems in healthcare, investigating the use of biomarkers and machine learning approaches to identify early conditions and avoid Burnout Syndrome. In this paper, a set of machine learning (ML) algorithms was evaluated using statistical data on biomarkers from the AffectiveRoad database to determine whether the use of explanations can help identify stress more objectively. This research integrates explainability and machine learning to aid in the identification of various levels of stress, which has not been previously evaluated for the domain of occupational stress. The Random Forest is the best-performing model for this assignment, followed by k-Nearest Neighbors and Neural Network. Later, explainers were applied to the Random Forest, highlighting feature importance, partial dependencies between characteristics, and a summary of the impact of features on outputs based on their values.
Download