In the June 2020 I had an article published ‘Affective computing in the modern workplace’ examining the growing use of emotion detection in data.
The full article can be accessed online in the Business Information Review journal. It is structured in three parts: First, introducing affect and affective computing, a phrase coined by computer scientist Rosalind Picard in 1995. Second, looking at affective computing in action – how the algorithms detect emotion in image and text using practical examples, and how the use cases align with Picard’s levels of affective computing. Third, reflecting on some considerations before use – the good and bad of trying to detect human emotions in real-world situations.
Picard defined levels of affective computing (Picard, 1995) and these can be used to evaluate the capabilities of affective computing solutions, from a complete absence of affect to a two-way interaction between a human and computer both recognising and expressing affect.
A key concern with the growing demand for using computers and machine learning to automatically detect emotions expressed in images and text is that we still do not yet have a robust definition for emotion, or what categories of emotion exist and whether or not they are distinct and/or universal (Izard, 2010). This becomes even more apparent when considering the role that emotion plays in cognitive decision-making. A popular hypothesis of how the human mind works is that it has three interdependent functions: cognition, affect and conation (Hilgard, 1980).
Much of cognitive science has focused on the cognition function – perceiving, thinking and intellect. But there is now growing recognition of the roles of affect and conation on cognition. Affect concerns our emotions, feelings and mood. Conation represents motivations, volition and temperament that also play a role in forming decisions.
For a long time, it was assumed that emotions were a problem, that they interfered with rational thought and led to poor decisions. However, neuroscientist Antonio Damasio studied the decision-making behaviour of patients with damage to areas of the brain associated with emotion and found that an absence of emotion could lead to an inability to make any decision at all (Damasio, 1994). In reality, we are faced with numerous decisions every day and often have to pick between seemingly arbitrary choices. Damasio proposed the ‘Somatic marker hypothesis’, stating that emotion is in the loop of reason and assists rather than disturbs rational choice. Less has been said about the role of conation in decision-making, beyond well-known demonstrations of risk-taking versus risk-averse behaviour. Biologists, anthropologists, geneticists and neuroscientists have all made, and continue to make, discoveries about variations in temperament, volition and motivation. A decision or action at any moment in time is a complex interplay between these three functions of the mind. To be able to reduce that complexity to a simple measurement of emotion based on an expression is, at best, a crude approximation.
Does it mean affective computing is impossible? The article goes into more detail… 🙂
Richardson S. (2020) Affective computing in the modern workplace. Business Information Review. Vol 37(2):78-85. doi:10.1177/0266382120930866
Damasio, A. (1994) Descartes’ Error: Emotion, Reason, and the Human Brain, Vintage edition, 2003.
Hilgard, E R. (1980) The trilogy of mind: cognition, affect and conation. Journal of the History of the Behavioural Sciences. Vol 16: 107-117.
Izard, C. (2010) The Many Meanings/Aspects of Emotion: Definitions, Functions, Activation, and Regulation. Emotion Review. Vol 2(4): 363-370.
Konner, M J. (1982) The Tangled Wing: Biological Constraints on the Human Spirit. First Owl Books edition, 2003
Picard, R. (1995) Affective computing. MIT Media Laboratory Perceptual Computing Section Technical Report No 321