An Introduction to Computational Affective Science (CAS)

What is CAS?

Affect refers to the fundamental neural processes that generate and regulate emotions, moods, and feeling states. Affect and emotions are central to how we organize meaning, to our behaviour, to our health and well-being, and to our very survival. Despite this, and even though most of us are all intimately familiar with emotions in everyday life, there is much we do not know about how emotions work, and how they impact our lives. Affective Science is a broad interdisciplinary field that explores these and other related questions about affect and emotions. Since language is a powerful mechanism of emotion expression, there is a growing use of language data and advanced natural language processing (NLP) algorithms to shed light on fundamental questions about emotions. I refer to this area of work as Computational Affective Science (CAS).

Why this webpage?

CAS is not widely known -- for example, many NLP researchers assume work on emotions is restricted to the automatic prediction of emotion categories. This is a lost opportunity. In recent years, I have often given talks on CAS work, and now I am creating this webpage as a small effort to further increase the understanding and visibility of this compelling area of research where NLP, ML, AI, Data Science, Neuroscience, Psychology, and Computational Social Science have much to contribute. The page will include:

• Basic information about what is CAS and how it is different from related fields
• Links to tutorials, workshops, survey articles, and talks on CAS
• Areas of research within CAS

Links to Tutorials, Workshops, Articles, and Talks on CAS

• Workshop

  Inaugural Workshop on Computational Affective Science. Christopher Bagdon, Krishnapriya Vishnubhotla, Kristen A. Lindquist, Lyle Ungar, Roman Klinger, and Saif M. Mohammad. LREC 2026, May 2026, Palma De Mallorca, Spain.

• Tutorials

NLP for Affective Science: Exploring Fundamental Questions on Emotions through Language and Computation. Krishnapriya Vishnubhotla and Saif M. Mohammad. IJCNLP-AACL 2025, Dec 23, 2025, Mumbai, India.
Slides
Video

A Tutorial on NLP for Affective Science: Exploring Fundamental Questions on Emotions through Language and Computation. Linz NLP Summer School, July 2024. Linz, Vienna.
Slides

• My work in CAS.
  
  

CAS and Other Areas of Research

CAS vs Sentiment Analysis (SA): While sentiment workshops and sentiment tracks have existed for almost two decades now, their focus has predominantly been on developing novel automatic algorithms for sentiment and emotion detection. On the other hand, CAS centesr the understanding of emotions and associated phenomena. A novel algorithm that helps achieve that goal is great, but not required. It makes more sense to use an approach that best addresses the research question---even if the computational approach is simple or well known. 

CAS vs. Affective Computing (AC): Affective Computing is an interdisciplinary  field with a focus on developing systems and devices that can recognize, interpret, and express human emotions (Picard 97). In contrast CAS has a focus on the science of how emotions work and how they impact our lives. 

CAS, SA, and AC have overlaps, but also clear differences, especially in terms of their purpose and focus. This also necessitates differences in how these works are reviewed. While SA and AC value algorithmic and computational novelty, CAS values new findings in the understanding of affect and emotions.

Interdisciplinary: A common critique of NLP work is the lack of theory. Good CAS work brings together ideas from NLP, Psychology, Cognitive Science, Social Science, etc. Simultaneously, practitioners of emotion/sentiment analysis and generation will benefit from an increased exposure to the latest theories and findings in Affective Science.

Data

All data associated with affect and emotions are relevant to CAS. Language data (text and speech) are especially interesting, but also relevant are other modalities such as facial and bodily expression, physiological signal processing (ECG, EEG, GSR, multimodal biosensing), etc. Multimodal data and data from various cultures from around the world are also of great significance.

SUBAREAS of CAS

A wide variety of work related to affect and emotions are part of CAS, including work on:

1. The Nature of Affect and Computational Modeling of Emotions Computational experiments that add to our understanding of affect and emotions. This includes findings relevant to theories of emotion, the biology of emotions, the neuroscience of emotions, and models of emotions such as appraisal models, dimensional models (valence / arousal / dominance), models of constructed emotion, cognitive-affective architectures, emotion dynamics (how emotions emerge, intensify, decay, or transition over time), emotion granularity, emotion regulation, affective embodiment, evolutionary affect development, developmental affect (how emotions and affect change over a life span), emotion and cognition, etc.  Note that many of these can be applied to:  human beings, animals, and even artificial agents.
   
   
2. Affective Data and Resources Work on compiling and annotating affect-related information in text, speech, facial and bodily expression, physiological signal processing (ECG, EEG, GSR, multimodal biosensing), etc. Since this proposal is for a workshop at an NLP conference, there will be a focus on text data (monolingual, multilingual) as well as multimodal data. Data from underserved languages is especially welcome.
   
   
3. Emotion Recognition, Prediction, and Inference At an Instance level (e.g., for each social media post):  Emotion classification (discrete emotions, dimensional ratings). Emotion intensity estimation. Emotion cause detection (what triggers an emotion in text, video, or interaction). Context-aware affect inference (considering culture, situation, social setting). At an Aggregate Level Creating emotion arcs from numerous utterances/posts Determining broad trends in emotion over time, across locations, towards entities of interest (e.g., climate chance), etc. Document-level and cross-document emotion analysis.
   
   
4. Applications Affect and Health, psychopathology, mental disorders  Affect and Behavior/Social Science (Modeling interpersonal affect, empathy, group-level affect modeling, polarization, affect contagion, computational models of emotion regulation) Affect and Education Affect and Literature/Storytelling/Digital Humanities Affect and Commerce
   
   
5. Explainability and interpretability in computational affective models.
   
   
6. Ethics, Fairness, Theory Integration, Philosophical Implications Bias and generalizability of affective systems across demographics. Privacy and ethics in affective data collection. Critically examining whether automatic NLP systems are relying on the current and valid theories of affect and emotion. What it means for machines to model or simulate affect. Broader societal implications of affective artificial agents.