A general theoretical framework for the design of artificial emotion systems in Autonomous Agents

Autonomous Agents (AAs) capable of exhibiting emotional behaviors have contributed to the development of natural human-machine interactions in several application domains. In order to provide AAs with emotional mechanisms, their underlying architecture must implement an Artificial Emotion System (AES), a computational model that imitates specific facets of human emotions. Although several AES have been reported in related literature, their design is generally supported on several emotion theories, leading researchers to model and integrate isolated emotion components and mechanisms into the architectures of AES. This theoretical foundation of AES contributes to ambiguities in the analysis and comparison of their underlying architectures, which demands the definition of standards, design guidelines, and integrative frameworks. In this paper, we present a psychologically inspired theoretical framework designed to serve as a platform for the unification of AES' components, the comparison of AES, and the design and implementation of AES in AAs. We analyze common emotion-related requirements of AES, emotion components involved in the design of this type of computational model, and emotion theories that drive the design of most AES. The validation of this framework demonstrates its compatibility with current AES and its feasibility as a model for unifying multiple emotional theories.     

Toward integrating cognitive components with computational models of emotion using software design patterns

Computational models of emotion (CMEs) are software systems designed to imitate particular aspects of human emotions. The main purpose of this type of computational model is to capture the complexity of the human emotion process in a software system that is incorporated into a cognitive agent architecture. However, creating a CME that closely imitates the actual functioning of emotions demands to address some challenges such as (i) sharing information among independently developed cognitive and affective components, and (ii) interconnecting complex cognitive and affective components that must interact with one another in order to generate realistic emotions, which may even affect agents’ decision making. This paper proposes an architectural pattern aimed at cataloging and describing fundamental components of CMEs and their interrelationships with cognitive components. In this architectural pattern, external cognitive components and internal affective components of CMEs do not interact directly but are extended by including message exchange methods in order to use a publish-subscribe channel, which enables their intercommunication, thus attenuating issues such as software heterogeneity. This structural approach centralizes communication management and separates the inherent complexity of the cognitive and affective processes from the complexity of their interaction mechanisms. In so doing, it enables the design of CMEs’ architectures composed of independently developed affective and cognitive components. The proposed architectural pattern attempts to make progress in capturing the complex process of human emotions in a software system that adheres to software engineering best practices and that incorporates quality attributes such as flexibility and interoperability.     

Better together: a unified perspective on appraisal and emotion regulation

ABSTRACT

Advances in our understanding of appraisal processes and emotion regulation have been two of the most important contributions of research on cognition and emotion in recent decades. Interestingly, however, progress in these two areas has been less mutually informative than one might expect or desire. To help remedy this situation, we provide an integration of appraisal theory and the process model of emotion regulation by describing parallel, interacting and iterative systems for emotion generation and emotion regulation. Outputs of the emotion generation system are perceived by the emotion regulation system, and emotion regulation strategies then modulate emotion by intervening at specific stages of the emotion generation system, ultimately changing appraisal dimensions. We hope that our unified perspective will encourage and guide future research at the interface of cognition and emotion.     

There's more to emotion than meets the eye: A processing bias for neutral content in the domain of emotional prosody

Research on emotion processing in the visual modality suggests a processing advantage for emotionally salient stimuli, even at early sensory stages; however, results concerning the auditory correlates are inconsistent. We present two experiments that employed a gating paradigm to investigate emotional prosody. In Experiment 1, participants heard successively building segments of Jabberwocky “sentences” spoken with happy, angry, or neutral intonation. After each segment, participants indicated the emotion conveyed and rated their confidence in their decision. Participants in Experiment 2 also heard Jabberwocky “sentences” in successive increments, with half discriminating happy from neutral prosody, and half discriminating angry from neutral prosody. Participants in both experiments identified neutral prosody more rapidly and accurately than happy or angry prosody. Confidence ratings were greater for neutral sentences, and error patterns also indicated a bias for recognising neutral prosody. Taken together, results suggest that enhanced processing of emotional content may be constrained by stimulus modality.     

A cognitive architecture for modeling emotion dynamics: Intensity estimation from physiological signals

Current approaches to emotion recognition do not address the fact that emotions are dynamic processes. This work concerns itself with the development of a cognitive architecture for modeling the dynamics of emotions with specific focus on a gray-box model for dynamic emotion intensity estimation that can incorporate findings from appraisal models, specifically Scherer’s Component Process Model. It is based on Dynamic Field Theory which allows the combination of theoretical knowledge with data-driven experimental approaches. A user study is conducted applying the proposed model to estimate intensity of negative emotions from physiological signals. Results show significant improvements of the proposed model to common methodology and baselines. The flexible cognitive architecture opens a wide field of experiments and directions to deepen the understanding of emotion processes as a whole.     

Socially emotional brain-inspired cognitive architecture framework for artificial intelligence

Artificial Intelligence of the next generation needs to interact with users socially, convincing them in its ability to understand human minds, including emotions. For this to happen, an artificial emotional intelligence is needed, capable of adequate, believable behavior in social emotional interactions. Building on previous developments, the present work extends the general framework of emotional Biologically Inspired Cognitive Architecture (eBICA: Samsonovich, 2013, 2018), endowing it with fluents describing, in addition to appraisals, somatic markers, feelings, emotions, moods, emotional reactions and biases. Key building blocks that integrate them are moral schemas and semantic maps. The model describes interaction of three factors: plans and commitments, moral and ethical values, and somatic comfort. Learning in this framework includes self-organization of semantic maps that in turn may provide guidance for active humanlike learning. Implications for empirical studies and practical applications are discussed together with the expected impact.     

The past thirty years of emotion research: appraisal and beyond

ABSTRACT

For this Special Issue, I highlight the past and present importance of appraisal theory as well as the challenges to its status as a total theory of emotions from the other functions of emotions: associative learning, self-regulation and social communication. This theoretical view applies both to emotion research in general and the specific fields of my interest in the emotions of moral judgment and intergroup processes. Methodologically, developments in analyses of large and more naturally occurring data sets will give an opportunity to square psychology’s structural models of discrete emotions with the more complicated reality that exists. Both for the field and for individual researchers picking up the study of emotions, my advice is to pay special attention to measures, their assumptions and their context.     

Implicit emotion regulation under demanding conditions: The moderating role of action versus state orientation

Action orientation is a volitional mode that promotes flexible self-regulation of emotional and motivational states; state orientation represents the conceptually opposite volitional mode that promotes fixation on (particularly negative) emotional and motivational states (Kuhl & Beckmann, 1994a). The present research investigated the link between action versus state orientation and implicit emotion regulation under demanding conditions. After inducing a demanding context, action-oriented participants displayed reduced affective priming effects of negative primes relative to state-oriented individuals (Studies 1–3). Action versus state orientation did not moderate affective priming effects of positive prime words (Studies 1–3). Effects of action versus state orientation emerged only for a limited number of trials (Study 2) and were reversed under low-demanding conditions (Study 3). These findings support the notion that implicit emotion regulation is closely linked to volitional action control.     

The semantic structure of emotion words across languages is consistent with componential appraisal models of emotion

Appraisal theories of emotion, and particularly the Component Process Model, claim that the different components of the emotion process (action tendencies, physiological reactions, expressions, and feeling experiences) are essentially driven by the results of cognitive appraisals and that the feeling component constitutes a central integration and representation of these processes. Given the complexity of the proposed architecture, comprehensive experimental tests of these predictions are difficult to perform and to date are lacking. Encouraged by the “lexical sedimentation” hypothesis, here we propose an indirect examination of the compatibility of the theoretical assumptions with the semantic structure of a set of major emotion words as measured in a cross-language and cross-cultural study. Specifically, we performed a secondary analysis of the large-scale data set with ratings of affective features covering all components of the emotion process for 24 emotion words in 27 countries, constituting profiles of emotion-specific appraisals, action tendencies, physiological reactions, expressions, and feeling experiences. The results of a series of hierarchical regression analyses to examine the prediction of the theoretical model are highly consistent with the claim that appraisal patterns determine the structure of the response components, which in turn predict central dimensions of the feeling component.     

On the dynamics of implicit emotion regulation: Counter-regulation after remembering events of high but not of low emotional intensity

Valence biases in attention allocation were assessed after remembering positive or negative personal events that were either still emotionally hot or to which the person had already adapted psychologically. Differences regarding the current state of psychological adjustment were manipulated experimentally by instructing participants to recall distant vs. recent events (Experiment 1) or affectively hot events vs. events to which the person had accommodated already (Experiment 2). Valence biases in affective processing were measured with a valence search task. Processes of emotional counter-regulation (i.e., attention allocation to stimuli of opposite valence to the emotional event) were elicited by remembering affectively hot events, whereas congruency effects (i.e., attention allocation to stimuli of the same valence as the emotional event) were obtained for events for which a final appraisal had already been established. The results of our study help to resolve conflicting findings from the literature regarding congruent vs. incongruent effects of remembering emotional events on affective processing. We discuss implications of our findings for the conception of emotions and for the dynamics of emotion regulation processes.     

A mechanism for biasing the appraisal process in affective agents

In this paper we present a mechanism to model the influence of agents’ internal and external factors on the emotional evaluation of stimuli in computational models of emotions. We propose the modification of configurable appraisal dimensions (such as desirability and pleasure) based on influencing factors. As part of the presented mechanism, we introduce influencing models to define the relationship between a given influencing factor and a given set of configurable appraisal dimensions utilized in the emotional evaluation phase. Influencing models translate factors’ influences (on the emotional evaluation) into fuzzy logic adjustments (e.g., a shift in the limits of fuzzy membership functions), which allow biasing the emotional evaluation of stimuli. We implemented a proof-of-concept computational model of emotions based on real-world data about individuals’ emotions. The obtained empirical evidence indicates that the proposed mechanism can properly affect the emotional evaluation of stimuli while preserving the overall behavior of the model of emotions.     

Toward a cumulative science of emotion: a functional-cognitive framework for emotion research

ABSTRACT

Definitions of emotion and emotional phenomena are often infused with intuitions and theoretical ideas about what is “truly” emotional. Although these intuitions and ideas motivate people to study emotion, their prominence at the conceptual level can hamper progress in emotion research. In this paper, we argue that there is merit in defining emotional phenomena as much as possible in terms of behavioural principles that have been developed outside of emotion research. We clarify that such a functional approach is compatible with, and can even strengthen, cognitive approaches to emotion research. This functional-cognitive perspective reveals ways to increase the cumulative nature of emotion research and to surpass initial intuitions and theoretical ideas.     

Implicit theories of emotion shape regulation of negative affect

Implicit theories of emotion—assumptions about whether emotions are fixed (entity theory) or malleable (incremental theory)—have previously been shown to influence affective outcomes over time. We examined whether implicit theories of emotion also relate to the immediate regulation of negative affect. Consistent with our hypotheses, we found that the more students endorsed an entity theory of emotion, the more discomfort they reported while watching an aversive movie clip, the more they avoided affective stimuli in this movie clip, the more negative affect they reported after the clip, and the less likely they were to watch the same clip again to learn about its ending. These findings suggest that implicit theories of emotion might produce poor affective outcomes immediately as well as over time. They also offer insight into why some people avoid negative affect while others confront it.     

The semantic pointer theory of emotion: Integrating physiology,appraisal, and construction

Emotion theory needs to explain the relationship of language and emotions, and the embodiment of emotions, by specifying the computational mechanisms underlying emotion generation in the brain. We used Chris Eliasmith’s Semantic Pointer Architecture to develop POEM, a computational model that explains numerous important phenomena concerning emotions, including how some stimuli generate immediate emotional reactions, how some emotional reactions depend on cognitive evaluations, how bodily states influence the generation of emotions, how some emotions depend on interactions between physiological inputs and cognitive appraisals, and how some emotional reactions concern syntactically complex representations. We contrast our theory with current alternatives, and discuss some possible applications to individual and social emotions.     

Cognition and emotion: a plea for theory

ABSTRACT

Research on cognition and emotion during the past 30 years has made reasonable progress in theory, methods and empirical research. New theories of the cognition–emotion relation have been proposed, emotion research has become more interdisciplinary, and improved methods of emotion measurement have been developed. On the empirical side, the main achievement of the past 30 years is seen to consist in the reduction of the set of serious contenders for a theory of emotions. Still, several important issues are not fully resolved, including the computational implementation of appraisal processes, the nature of emotions, and the link between emotions and actions. Also, quantitative theories of the cognition–emotion relation need to be refined and tested, and improved theories of the link between emotions and bodily and facial expressions need to be developed. To counter the dangers of theoretical fragmentation and knowledge loss, more efforts should be devoted to the analysis, reconstruction, comparison and integration of important theories and hypotheses in the field of emotion, as well as to the systematization of arguments in favor and against these theories and hypotheses.     

Investigating a process model for leader affective presence,interpersonal emotion regulation,and interpersonal behaviour in teams

Leader affective presence is the tendency of leaders to elicit feelings that are consistent among other individuals, and has been supported as a relevant personality trait for understanding teamwork. Drawing on a model that integrates personality and emotion regulation, this study aimed to expand research on affective presence by proposing team members’ perceptions of leader interpersonal emotion regulation as a process that explains how leader affective presence is related to team member behaviour. In the model, teamness—the perception that interdependence and reflexivity are required in the team—is presented as a boundary condition to the effects of affective presence via emotion regulation. Results of a study conducted with 99 teams showed that team member ratings of leader positive affective presence were linked to their perceptions that leaders had used affect-improving emotion regulation which in turn was associated with greater team citizenship behaviour. Contrariwise, team member ratings of leader negative affective presence were associated with perceived use of affect-worsening emotion regulation by leaders which in turn was associated with lower levels of team citizenship, but only when teams were low in teamness. These findings contribute to understanding how leaders’ individual differences are related to teamwork through affective processes.     

What the face and body reveal: in-group emotion effects and stereotyping of emotion in African American and European American children

This study examined whether 3- to 7-year-old African American and European American children’s assessment of emotion in face-only, face + body, and body-only photographic stimuli was affected by in-group emotion recognition effects and racial or gender stereotyping of emotion. Evidence for racial in-group effects was found, with European American children being more accurate when assessing emotion in European American photographs than African American photographs for some emotions. African American children were either equally proficient in recognizing emotion in African American and European American photographs or were more accurate with European American photographs for some emotions. Stereotyping of emotion was also found, with boys being more often labeled with “masculine” emotions (e.g., mad) and at least some girls being more often labeled with “feminine” emotions (e.g., happy). However, stereotyping effects were found only when the face was present in the stimuli and were not found with body-only stimuli. In-group effects, however, were not affected by type of photograph (face-only, body-only, or face + body), with children being unable to recognize at least some emotions from just the body postures alone (mad). These results have important implications for how future studies assess emotion recognition in children, particularly in terms of how emotion stimuli are constructed, the diversity of the stimuli, and who judges the stimuli.     

The roles of leader emotion management and team conflict for team members' personal initiative: A multilevel perspective

Employees' personal initiative (i.e., their active and future-oriented engagement at work) is of high relevance to organizations that strive for innovation and competitiveness. To better understand its drivers and diminishers, the present study refers to Affective Events Theory and examines the impact of leader emotion management, team conflict, and affective well-being on personal initiative. Data from 300 members of 59 work teams in Germany were gathered from multiple sources at three points in time over the course of 2 weeks. Multilevel analyses indicated that leader emotion management positively affected team members' personal initiative and that this effect was mediated by team members' affective well-being. Data also revealed a contrasting negative indirect effect of task conflict within the team on team members' personal initiative via team members' affective well-being. The study's results suggest that the improvement of leaders' emotion management as well as the establishment of work conditions and experiences that foster affective well-being should be seriously considered by organizations wishing to facilitate personal initiative in team settings.     

Fear and perceived uncontrollability of emotion: Evaluating the unique contribution of emotion appraisal variables to prediction of worry and generalised anxiety disorder

Novel theoretical frameworks place the symptom profile of generalised anxiety disorder (GAD) within the context of dysfunctional emotional processes. It is suggested that fear and intolerance of emotions exacerbate subjective distress and motivate the use of maladaptive coping strategies, such as worry. To date, studies evaluating these models have suffered two key limitations. Firstly, few studies have involved treatment-seeking samples, and secondly, none have evaluated the unique variance attributable to emotion appraisal variables above and beyond previously established predictors of worry and GAD. The present study begins to address these limitations by assessing the contribution of fear and perceived uncontrollability of emotions in predicting worry and clinical GAD status after controlling for variance attributable to depressive symptoms, meta-cognitive beliefs, intolerance of uncertainty, and perceptions of external threat. Supporting current models, results showed that perceived control over emotional reactions was a unique predictor of GAD diagnostic status and both clinical and non-clinical worry.