Volume 5, Issue 2 (2025)
jpt 2025, 5(2): 207-221 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Mashhadi Raviz F, Nabavi L, Alizadeh M. Logical Formalization of Some Counterfactual Emotions. jpt 2025; 5 (2) :207-221
URL: http://jpt.modares.ac.ir/article-34-80898-en.html
URL: http://jpt.modares.ac.ir/article-34-80898-en.html
1- Department of Philosophy and Logic, Faculty of Humanities, Tarbiat Modares University, Tehran, Iran
2- Department of Computer Science, School of Mathematics, Statistics and Computer Science, College of Science, University of Tehran, Tehran, Iran
2- Department of Computer Science, School of Mathematics, Statistics and Computer Science, College of Science, University of Tehran, Tehran, Iran
Full-Text [PDF 766 kb]
(161 Downloads)
| Abstract (HTML) (298 Views)
Shame
Another important self-conscious emotion that contributes to agent behavior in MAS is shame. Shame is often viewed as a public emotion, rooted in an awareness of how others perceive one’s actions. Such awareness is indicative of social judgments on the part of an agent and can trigger feelings of unworthiness or inadequacy. In MAS, the impetus for decision and behavioral derivation is the elicitation of shame, which dominates agents because, sometimes, to conform to social norms, disapproval must be avoided [Covert et al., 2003; Adam et al., 2013].
Shame typically arises when an agent knows that others will perceive one’s actions as inappropriate but desires to overcome that negative perception. In the CFE framework, shame emerges when an agenti a r φ i j i i j
Action-Performance: The agent performed the actiona 
Awareness of Others’ Knowledge: Agenti j i j 
Knowledge of Counterfactual Avoidability: The agent knows that it could have avoidedφ 
Desire for positive judgment: The agent desires to overcome the negative judgment.
Full-Text: (10 Views)
Introduction
The main challenge in AI is creating cognitive interactive systems that exhibit human-like behaviors, actions, and emotions, thereby qualifying as humanoid agents. Thus, a humanoid agent must be able to think about emotions, represent its own emotions and personality, attribute emotions to humans, predict the impact of its actions on human emotions, and change or modify its behavior accordingly. So, the study of emotional phenomena is an important area in AI for creating the next generation of emotional interaction systems. Emotional computing has been growing in recent years. Some are interested in creating logical frameworks for the formal analysis of emotions and are mainly focused on using logical methods to give artificial agents precise specifications on how to represent their emotional states.
Notable prior work includes the logical model for counterfactual emotions by Lorini and Schwarzentruber (2011), which concentrated on the interplay between choices, abilities, the consequences of agent and group actions, knowledge, and desires. Steunebrink et al. (2012) provided a logical representation of a segment of a well-known psychological model of emotions, systematically examining the hierarchical conditions under which emotions are triggered, and applying these findings to guide the analysis of emotional stimuli [Steunebrink et al., 2012]. Furthermore, Guiraud et al. (2011) integrated emotion theory with speech act theory and logic, employing concepts of belief, goal, ideality, and responsibility within modal logic to elucidate what a system communicates in dialogue with another system [Guiraud et al., 2011].
Counterfactual Emotions Framework
In this paper, we have proposed a logical framework for counterfactual emotions (CFE) within MAS. In CFE, we have included the essential moral emotions guilt, shame, as well as capturing their distinct characteristics and roles with respect to moral and social reasoning. Guilt refers to the recognition by the agent that he or she has violated a norm, while shame follows from awareness of the judging disapproving eye of others.
This framework is built upon finite sets of normative propositions (Norm) that define expected behaviors and evaluation criteria (EC) that set standards for desirable agent characteristics. By incorporating new atomic constructs such as
(action performance), 
(action leading to an outcome), 
(immorality of an outcome), 
(inadequacy related to an outcome), and 
(counterfactual avoidability of an outcome), CFE provides the tools to precisely define complex moral and evaluative concepts.
The main challenge in AI is creating cognitive interactive systems that exhibit human-like behaviors, actions, and emotions, thereby qualifying as humanoid agents. Thus, a humanoid agent must be able to think about emotions, represent its own emotions and personality, attribute emotions to humans, predict the impact of its actions on human emotions, and change or modify its behavior accordingly. So, the study of emotional phenomena is an important area in AI for creating the next generation of emotional interaction systems. Emotional computing has been growing in recent years. Some are interested in creating logical frameworks for the formal analysis of emotions and are mainly focused on using logical methods to give artificial agents precise specifications on how to represent their emotional states.
Notable prior work includes the logical model for counterfactual emotions by Lorini and Schwarzentruber (2011), which concentrated on the interplay between choices, abilities, the consequences of agent and group actions, knowledge, and desires. Steunebrink et al. (2012) provided a logical representation of a segment of a well-known psychological model of emotions, systematically examining the hierarchical conditions under which emotions are triggered, and applying these findings to guide the analysis of emotional stimuli [Steunebrink et al., 2012]. Furthermore, Guiraud et al. (2011) integrated emotion theory with speech act theory and logic, employing concepts of belief, goal, ideality, and responsibility within modal logic to elucidate what a system communicates in dialogue with another system [Guiraud et al., 2011].
Counterfactual Emotions Framework
In this paper, we have proposed a logical framework for counterfactual emotions (CFE) within MAS. In CFE, we have included the essential moral emotions guilt, shame, as well as capturing their distinct characteristics and roles with respect to moral and social reasoning. Guilt refers to the recognition by the agent that he or she has violated a norm, while shame follows from awareness of the judging disapproving eye of others.
This framework is built upon finite sets of normative propositions (Norm) that define expected behaviors and evaluation criteria (EC) that set standards for desirable agent characteristics. By incorporating new atomic constructs such as
(action performance), (action leading to an outcome), (immorality of an outcome), (inadequacy related to an outcome), and (counterfactual avoidability of an outcome), CFE provides the tools to precisely define complex moral and evaluative concepts.Norm and EC are sets of propositions within the logical framework, they differ fundamentally in their purpose, nature, and role in shaping agent behavior and emotions: Norms are prescriptive, meaning they specify what agents should or should not do. They function as external constraints on behavior, and their violation is formally linked to the concept of immorality. Such violations are critical triggers for counterfactual emotions, particularly guilt. Unlike norms, evaluation criteria are descriptive rather than prescriptive. They do not dictate behavior but serve as criteria for evaluating an agent’s state or performance. Failure to satisfy an evaluation criterion is captured through the concept of inadequacy that often correlates with emotions such as shame.
After a broad introduction to the syntax and language of CFE, we presented its semantics. Then, the axioms of CFE were introduced, followed by the statement and proof of two important theorems: soundness and completeness.
Counterfactual emotions
Counterfactual emotions are some of the most important psychological phenomena in which an agent comments on how seemingly different choices would have changed the course of events. They include guilt and shame, but are also quite relevant in dictating one’s behavior in and toward social contexts, in decision making, and in shaping agents’ actions in MAS. This cognitive mechanism allows agents to learn from their experiences, even the bitter ones of the past, but with the effect of generating emotional responses that motivate behavior adjustment [Roese et al., 2009].
Understandably, counterfactual emotions evoke guilt and shame, and these are most visible in MAS. Agents feel guilty because they realize that their actions violate personal or social standard norms, so it is immediately appealing to them to perform reparative behaviors. Shame, on the other hand, has to do with self-comparative evaluation and judgment of others, which usually leads to withdrawal or avoidance behavior. Thus, these emotions can also promote the ability to better simulate agent behavior by designing systems that are properly equipped with human ethical considerations.
Guilt
Guilt is one of the self-conscious emotions that depends on self-awareness for its emergence. Self-conscious emotions are directly related to the self-concept of evaluation and arise not from the stimulus but from the cognitive ability to scrutinize oneself with moral reasoning or to evaluate socially meaningful consequences [Tangney et al., 2007].
Guilt is generally understood as a subjective kind of internal attribution for having violated norms or caused harm to others. Within the CFE framework, guilt arises when an agent.i a r φ i
Action-Performance: Agenti a 
Knowledge of Immorality: Agenti φ 
Knowledge of Counterfactual Avoidability: The agent knows that it could have avoidedφ 
After a broad introduction to the syntax and language of CFE, we presented its semantics. Then, the axioms of CFE were introduced, followed by the statement and proof of two important theorems: soundness and completeness.
Counterfactual emotions
Counterfactual emotions are some of the most important psychological phenomena in which an agent comments on how seemingly different choices would have changed the course of events. They include guilt and shame, but are also quite relevant in dictating one’s behavior in and toward social contexts, in decision making, and in shaping agents’ actions in MAS. This cognitive mechanism allows agents to learn from their experiences, even the bitter ones of the past, but with the effect of generating emotional responses that motivate behavior adjustment [Roese et al., 2009].
Understandably, counterfactual emotions evoke guilt and shame, and these are most visible in MAS. Agents feel guilty because they realize that their actions violate personal or social standard norms, so it is immediately appealing to them to perform reparative behaviors. Shame, on the other hand, has to do with self-comparative evaluation and judgment of others, which usually leads to withdrawal or avoidance behavior. Thus, these emotions can also promote the ability to better simulate agent behavior by designing systems that are properly equipped with human ethical considerations.
Guilt
Guilt is one of the self-conscious emotions that depends on self-awareness for its emergence. Self-conscious emotions are directly related to the self-concept of evaluation and arise not from the stimulus but from the cognitive ability to scrutinize oneself with moral reasoning or to evaluate socially meaningful consequences [Tangney et al., 2007].
Guilt is generally understood as a subjective kind of internal attribution for having violated norms or caused harm to others. Within the CFE framework, guilt arises when an agent.
Action-Performance: Agent
Knowledge of Immorality: Agent
Knowledge of Counterfactual Avoidability: The agent knows that it could have avoided
Shame
Another important self-conscious emotion that contributes to agent behavior in MAS is shame. Shame is often viewed as a public emotion, rooted in an awareness of how others perceive one’s actions. Such awareness is indicative of social judgments on the part of an agent and can trigger feelings of unworthiness or inadequacy. In MAS, the impetus for decision and behavioral derivation is the elicitation of shame, which dominates agents because, sometimes, to conform to social norms, disapproval must be avoided [Covert et al., 2003; Adam et al., 2013].
Shame typically arises when an agent knows that others will perceive one’s actions as inappropriate but desires to overcome that negative perception. In the CFE framework, shame emerges when an agent
Action-Performance: The agent performed the action
Awareness of Others’ Knowledge: Agent
Knowledge of Counterfactual Avoidability: The agent knows that it could have avoided
Desire for positive judgment: The agent desires to overcome the negative judgment.
Conclusion
This study establishes an initial framework for developing more advanced intelligent systems. More specifically, it contributes to the effort of making AI agents capable not only of taking in information and performing tasks, but also of understanding, learning, and ultimately acting according to the complex ethical principles and social norms that define human societies.
This study establishes an initial framework for developing more advanced intelligent systems. More specifically, it contributes to the effort of making AI agents capable not only of taking in information and performing tasks, but also of understanding, learning, and ultimately acting according to the complex ethical principles and social norms that define human societies.
Article Type: Original Research |
Subject:
Logic and Philosophy of Logic (Analytical)
Received: 2025/04/23 | Accepted: 2025/05/28 | Published: 2025/06/4
Received: 2025/04/23 | Accepted: 2025/05/28 | Published: 2025/06/4
References
1. Covert MV, Tangney JP, Maddux JE, Heleno NM (2003). Shame-proneness, guilt-proneness, and interpersonal problem solving: A social cognitive analysis. Journal of Social and Clinical Psychology. 22(1):1-12. [Link] [DOI:10.1521/jscp.22.1.1.22765]
2. Guiraud N, Longin D, Lorini E, Pesty S, Rivière J (2011). The face of emotions: A logical formalization of expressive speech acts. Proceedings of the 10th International Conference on Autonomous Agents and Multiagent Systems. Taipei: AAMAS. p. 1031-1038. [Link]
3. Lorini E, Schwarzentruber F (2011). A logic for reasoning about counterfactual emotions. Artificial Intelligence. 175(3-4):814-847. [Link] [DOI:10.1016/j.artint.2010.11.022]
4. Roese NJ, Morrison M (2009). The psychology of counterfactual thinking. Historical Social Research. 34(2):16-26. [Link]
5. Steunebrink BR, Dastani M, Meyer JC (2012). A formal model of emotion triggers: An approach for BDI agents. Synthese. 185(Suppl 1):83-129. [Link] [DOI:10.1007/s11229-011-0004-8]
6. Tangney JP, Stuewig J, Mashek DJ (2007). Moral emotions and moral behavior. Annual Review of Psychology. 58:345-372. [Link] [DOI:10.1146/annurev.psych.56.091103.070145]
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |