Anguish in the Context of Affective Computing (emotion-ai)
Is it possible to model anguish computationally?
Anguish is an affection defined in the psychoanalysis field by Freud and by Lacan. A common translation for the word anguish from other languages such as French and German to English is anxiety. The probable reason for that will be explained later in this text.
Affective Computing
In case you want some context of what is affective computing or the taxonomy of approaches to modeling emotions in artificial systems (emotion-AI), you can find a high level overview in this link.
Let’s go back to the Anguish
After this high level overview of the emotion models, let’s go back to anguish. The motivation of that topic stems from social distance occurred in 2020 which was triggered by COVID-19 due to the fear of the unknown. It was not clear the effective ways of treatment for that new disease. This fear of the unknown motivated a study of this emotion, which appears in the Ekman’s model.
Fear is one of the seven universal emotions experienced by everyone around the world. Fear arises with the threat of harm, either physical, emotional, or psychological, real or imagined. While traditionally considered a “negative” emotion, fear actually serves an important role in keeping us safe as it mobilizes us to cope with potential danger. (What is Fear? by Paul Ekman)
Studies on the fear led to the philosopher Kierkegaard wrote a book about fear called Der Begriff Angst in German. The word Angst in Germany has multiple meanings; among them it is found anguish, anxiety and fear. It is possible that this originated the title The Concept of Anxiety in English and Conceito da Angústia (The Concept of Anguish) in Portuguese.
What is the difference between anguish, anxiety and fear?
Fear by definition has 2 states — fight and flight — and it is one of the basic universal emotions of Ekman’s model. Fear is an emotion that is also associated with System 1 (i.e., fast mode of thought) as presented by Daniel Kahneman in Thinking, Fast and Slow. Conversely, the definition of Anguish is “severe mental or physical pain or suffering.” And Anxiety, on the other hand, is “the state of feeling nervous or worried that something bad is going to happen”.
Fear and anxiety can be differentiated in four domains: (1) duration of emotional experience, (2) temporal focus, (3) specificity of the threat, and (4) motivated direction. Fear is short-lived, present-focused, geared towards a specific threat, and facilitating escape from threat; anxiety, on the other hand, is long-acting, future-focused, broadly focused towards a diffuse threat, and promoting excessive caution while approaching a potential threat and interferes with constructive coping.(WIKIPEDIA)
In the field of philosophy, Kierkegaard and Heidegger attributed anguish to nothing. According to Kierkegaard the nothingness was related to existentialism and while for Heidegger the nothingness was ontological. In the field of psychiatry, anguish is under anxiety disorder, which is explained in the DSM-5. In psychoanalysis, anguish is considered something existential and that was one of the matters Freud and Lacan dedicated a book for, they both believed that anguish is affection.
The book X of Lacan
At this point, it is pertinent to quote a little more about Lacan and Freud, since both dedicated studies on anguish. The book X of Lacan that handles the anguish has an English translation under the title Anxiety: The Seminar of Jacques Lacan, Book X; the French version has the title: Le Séminaire Livre X: L’angoisse; and the Portuguese O seminário, Livro 10: A Angústia.
It seems to be some evidences of a relation between Fear, Anxiety and Anguish, mainly because of the last two words: Anxiety and Anguish. It is possible to notice that the German word “Angst” has multiple meanings and that some were translated as “Anxiety” in English. Added to this, the French word “Angoisse” is also being translated as “Anxiety” in English versions. Lacan argues that Anguish is the lack of the lack, and it can only be cured with the desire.
By researching the content of Lacan’s Book X, it is possible to find an attempt to formally model psychoanalysis. Examples of this formalization are observed in the graph of desire (Figure 3), the optical schema of Lacan (Figure 4 and Figure 5), matheme among others. Following are some examples found in Book 10 of the way Lacan proposes formalizing the psychoanalysis:
Lacan uses graph, optical schema and formulas to formalize the psychoanalysis. Figures 4 and 5 show the optical schema proposed by Lacan with the concave mirror and plane mirror to explain the real and the imaginary of an individual perception.
Is it possible to apply the formalization of psychoanalysis developed by Lacan to the area of computing in order to create computational models that involve individual deepening knowledge of each human being?
Psychoanalytical Model
The article The Current State of Psychoanalytically-Inspired AI describes the idea of having a psychoanalytical model. It refers to Id, Ego and Super-ego which are terms that Freud brought to the psychoanalysis. Beyond those mentioned it proposes a hierarchical system with 3 layers as shown in the image below (Figure 6). The last layer is the Psychoanalytical Model (Psyche) which has instances of Ego, Id, and Super-Ego. The sub-functions can be viewed as part of a distributed system.
The article herein presents the concepts of psychoanalytic used to describe the input value that will trigger an action, it is related to the sub-functions, see table below.
Transference
An interesting concept in the area of mental health is transference , which by definition is a phenomenon that projects another into another. And that leads to another concept which is the individual perception that each one has about the world. And then, it leds to Lisa Feldman Barrett on How Emotions are Made which argues against about the classic way of emotion detection based on standard patterns.
An interesting concept in the area of mental health is transference , which is a phenomenon that a person projects their feelings or desires into another. This phenomenon leads to a concept that considers the individual perception that each person has about the world. Also, the neuroscientist Lisa Feldman Barrett who wrote How Emotions are Made argues against the classical way of emotion detection based on standard patterns.
How emotions are made
As mentioned in the How Emotions are Made(Cinematic Lecture) “One challenge comes from the fact that the same exact face can be used in different emotions” (BARRET, L. F., 2020) represented in Figure 7 and “Same emotions can have many faces”(BARRET, L. F., 2020) represented in Figure 8. The idea is that the same facial expression can have different meanings in different contexts; in addition, it argues that the brain guesses automatically and naturally by making predictions according to each individual’s experience. Therefore, the person can guess in a right or wrong way what another person is facing. The author says
“No one detects emotion in anyone else”, “your brain is predictive not reactive”, “Affect is not emotion, it is more basic than emotion, it exists because one of your brain’s most important job is running a budget for your body”, “Your brain is using predictions which become what you see to make sense of your affect”, “Concepts are predictions from our past experiences” and “Emotion are created from concepts which are the predictions that give meaning to your affect in your environment”. (BARRET, L. F., 2020)
Lisa’s research can be linked in some way to the concepts of transference and perception, as both use past experiences to help making some kind of decision; also given that it infers how another person feels by our own prior experience.
Therefore, it seems interesting that affective computing could be able to model emotions in artificial intelligence systems based on the perceptual lens of each individual. This perspective should be able to explain what really goes through the person via Explainable AI algorithms (X-AI). To date, the FACS explains why the image or video were classified within a certain category of emotion from the Ekman’s model, but after Lisa has demonstrated that the same facial expression can fit in different labeled meanings depending on situations, the idea to understand each individual lens could decrease the bias. Would it be possible to have machines detecting individuals’ emotions instead of emotions perception? Is it possible to create a system to help people use Lacan’s formalization of psychoanalysis?
Since Anguish is an affection, could we have an emotion-AI system to help manage our body budget?
Modeling anguish computationally
Based on what I have researched on this subject, the initial idea of building a model able to understand anguish is to design a psychoanalytical model which could use the theory behind psychology and psychoanalysis, especially with the work that Lacan has left for us but it would require having each individual’s inputs separately. Maybe in future we could identify person’s anguish automatically but at the present moment this would be like a reinforcement learning because it would depend on the individuals perception input.
References
BARRETT, F. L. Available at: https://how-emotions-are-made.com/notes/Home
BARRETT, F. L. How Emotions are Made. United States. HMHCO, 2017.
LACAN, J. O Seminário, Livro 10, a angústia [1962–1963]. Rio de Janeiro: Jorge Zahar, 2005.
LACAN, J. 2014. Anxiety: The Seminar of Jacques Lacan, Book X, ed. Jacques-Alain Miller. Trans. A. R. Price. Cambridge: Polity Press.
KAHNEMAN, D. Thinking, Fast and Slow. Farrar, Srauss and Groux / New York, 2011.
D. DIETRICH, S. SCHAAT, D. BRUCKNER, K. DOBLHAMMER and G. FODOR, The current state of psychoanalytically-inspired AI, IECON 2013–39th Annual Conference of the IEEE Industrial Electronics Society, Vienna, 2013, pp. 6666–6671, doi: 10.1109/IECON.2013.6700235.
