This is an article relative to a tricky, very delicate topic. The new frontier of the Artificial Intelligence in Videogames and Simulations goes toward a greater consciousness of the complexity of the human and animal behaviour. So far, very few videogames tried to use advanced AI models for defining the NPC behaviour (NPC stands for Non Playable Characters, that are alla the characters in a videogame not managed by players), and even in those cases, the distance of the results respect to a real human behaviour was higher than we would mention.
Fortunately, in the last years new AI models are appearing thanks to a greater confidence with the psychological studies. The subject of this set of articles is a new proposal for making better Intelligent Agents (used as NPCs for videogames), working on their ability to feel emotions in a similar way to humans and animals. New studies are pointed out the fact that emotions are the main factor of the human cognitive system that current AI systems don't have. This is then the new frontier of the Artificial Intelligence science, and the new videogames (and new human simulations systems) will largely take benefit of them.
This article try to help in making a new point of view about the possible AI models for simulating the psychological models of emotions.
This article try to help in making a new point of view about the possible AI models for simulating the psychological models of emotions.
In the last decades several psychologists made up some, different models with which try to reproduce the human (and animal) emotions. They started, often, from different approaches, like the Cognitive Appraisal theory (Lazarus and others, 1970), the anatomical point of view (Damasio A., 1994)(LeDoux J.E., 1996 and 2000),the Bi-Polar Dimension theory (Russell, 1980), the Three-Dimension emotions (Schlosberg, 1954) without mentioning old theories like the James-Lange’s and the Cannon-Bard’s.
Although all these theories come from scientific evidence, there is not a common point of view. There are at least two main approaches completely opposite one each other:
1) Discrete Emotions: the group of theories stating that emotions have to be discrete, differentiated among them by emotional reactions (Lazarus and others, 1970), in the body (excitatory phenomena) (Arnold, 1950) and how they arise (Lazarus and others, 1970).
2) Continuous Emotions: the group of theories stating that emotions can’t be seen as discrete labels (Russell, 1980, Schlosberg, 1952).
Both approaches are able to claim issues for the other one. As remembered by Basile (2012), the discrete emotions approach fails to explain phenomena such as the frequent comorbidity observed between different psychological disorders, nor does it solve the vexed question of the correspondence between emotions and a specific neurophysiological substrate. A criticism has moved for both approaches by anthropologists and scientists who study the differences of perception and reaction to emotions for the different human cultures.
Although all these theories come from scientific evidence, there is not a common point of view. There are at least two main approaches completely opposite one each other:
1) Discrete Emotions: the group of theories stating that emotions have to be discrete, differentiated among them by emotional reactions (Lazarus and others, 1970), in the body (excitatory phenomena) (Arnold, 1950) and how they arise (Lazarus and others, 1970).
2) Continuous Emotions: the group of theories stating that emotions can’t be seen as discrete labels (Russell, 1980, Schlosberg, 1952).
Both approaches are able to claim issues for the other one. As remembered by Basile (2012), the discrete emotions approach fails to explain phenomena such as the frequent comorbidity observed between different psychological disorders, nor does it solve the vexed question of the correspondence between emotions and a specific neurophysiological substrate. A criticism has moved for both approaches by anthropologists and scientists who study the differences of perception and reaction to emotions for the different human cultures.
Going in the opposite way respect to the other scientists, Roseman (1996), with his psychological model, shown that appraisal information “can vary continuously, but categorical boundaries determine which emotion will occur”. This is a point in favour of a commune representation of emotions, both continuous and discrete. Scherer (1984a) wanted to affirm that the distinction between continuous and discrete approach should be resolved by placing discrete emotional categories (i.e. happiness, sadness, etc.) while continuous models represent the varieties, styles, and levels of these already defined distinct emotions. Both Roseman’s and Shrerer’s point of view are something like saying that “outside emotions, they are seen as discreet, but inside they work in a continuum space”. This idea reminds me to the great diatribe between the classical physics and the quantic one, where the physical subatomic elements react as particle and/or as a wave.
I think that this point of view, the one with which we consider emotions having two different kinds of natures, one discreet and one continuous, strictly depends on the way with which we use or measure them. I suspect that this solution could be the right one to have better results in the AI implementation of the human and animal behaviours. Indeed, we need to remember that, if there is a continuous space between the level of arousal and the valence (positivity or negativity) for an emotion represented by the circumplex theory, the human reaction to emotions cannot have continuous values, but discreet kinds of reactions. This is the real problem between the two approaches: they see the same phenomena, but one (the continuous approach) sees it from its inside, relatively to its nature. The other (the discrete approach) sees it from outside, relatively to the consequences of its values and then the events/thoughts whom arose them as much as the actions made as a consequence of them.
Remembering that emotions are fundamentals for humans (and animals too) for taking decisions (Ambady N, Gray HM. 2002, Barsade SG. 2002, Bechara A, Damasio H, Damasio AR, Lee GP. 1999 only to mention some), and also that humans and animals take decision for making any sort of (more or less) reasoned reaction to an external event, we can say for sure that it’s not possible to face with a continuous space of possible emotions when we have to do with events. This is an impassable limit of the continuous approach in its generation (from events) and in its goal (the decision of a reaction). Contrarily, it’s not possible to state precisely which kind of emotion can be triggered by an event associated to a third individual who elicits a specified sentiment into the agent, without using a continuous approach.
As evolutionary psychology stated (and other psychologists agree with them), emotions are the result of the organisms evolution, real mind programs (or modules) able to trigger physical and mental responses in a way strictly dependent to the meaning of the event/thought whom arose them. The Arousal, the Valence and the Pleasure/Pain binomial are the means with which our mind triggers emotions. The natural evolution of species promoted the use of specified ranges of Arousal/Valence and Pleasure/Pain value to better predispose the organism to face with each particular event.
This means, for me, that both discrete and continuous approach are to use in the proper way and in the right context.
References
Barsade SG. 2002. The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly 47:644-75
Basile B. (2012). Un modello dimensionale delle emozioni: integrazione tra le neuroscienze dell’affettività, lo sviluppo cognitive e la psicopatologia. Cognitivismo.com
Bechara A, Damasio H, Damasio AR, Lee GP. (1999). Different contributions of the human amygdala and ventromedial prefrontal cortex to decision-making. Journal of Neuroscience 19:5473-81
Damasio A. (1994). Descartes' Error: Emotion, Reason, and the Human Brain, Putnam Publishing.
Lazarus, R. S., Averill, J. R., Opton, E. M. Jr. (1970). Toward a cognitive theory of emotions. In Feelings and Emotions, ed. M. Arnold, pp. 207-32. New York: Academic
LeDoux J.E .(1996). The Emotional Brain. New York: Simon and Schuster.LeDoux JE (2000). Emotion circuits in the brain. Annu Rev Neuroscience 23:155-184.
Russell, J. (1980). A circumplex model of affect Journal of Personality and Social Psychology 39: 1161–1178. doi:10.1037/h0077714
Scherer, K.R. (1984). On the nature and function of emotion: a component process approach. In Klaus R. Scherer and Paul Ekman (Ed.), Approaches to emotion (pp. 293–317). Hillsdale, NJ: Erlbaum.
Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review 61: 81–8. doi: 10.1037/h0054570
Schlosberg, H. (1952). The description of facial expressions in terms of two dimensions. J. exp. Psychol., 1952, 44, 229-237
