Quick: What mood are you in right now? This (loaded) question is interesting because it can help us deeply understand ourselves. It’s not that one’s current mood is highly informative, but that in order to understand oneself it helps to have a helpful mental model of moods. This is something I struggled with until I read Robert Thayer’s Calm Energy, and later, James Russell’s theory of “core affect”.
James Russell and Sylwia Hyniewska at ISRE-2017 (St-Louis, MO). July 29, 2017
At the biannual meeting of the International Society for Research on Emotion in St-Louis in July (ISRE-2017), Dr. Sylwia Hyniewska and I met with Dr. James Russell to discuss his theory of affect in relation to the paper Hyniewska and I had presented (on “emotion” and insomnia).
According to Russell’s theory, whereas emotions come and go, the conscious human mind is always at a specific point in a two-dimensional affective space. (See his 2009 and 2003 papers). Russell’s two dimensions of affect are pleasure-displeasure and activation-deactivation. This circumplex model is said to be a geometric rotation of Thayer’s model of moods, whose two dimensions are tension and energy. (I see some differences, but that is for another article.) For instance, elation is a high pleasure, high activation mental state. If you are distressed, you have high activation, low pleasure (high displeasure). You might not be reflectively aware that you are in a certain mood; or you might be aware of your mood but unable to describe it (see Annette Karmiloff-Smith, Beyond Modularity). But your mood is part of your conscious state.
I don’t know if anyone has ever pointed this out, but it strikes me that there is a similarity between Immanuel Kant’s very deep conception that time and space (deeply) structure our experience and Russell’s assumption that moods color our experience.
Russell postulates another difference between core affect and emotions: core affect doesn’t necessarily have a mental object. If you are angry (as an emotion), then odds are you are angry at someone. If you are in a truculent mood, however, you might not have an idea of the cause or target of your truculence. To distinguish moods from emotions, however, one must invoke a theory of emotions. Russell makes a big point of arguing that emotion is not a scientifically viable category. So, he doesn’t provide such a theory; but he could easily point to particular theories of “emotion” to substantiate his distinction between moods and emotions.
At this point you might be wondering whether Russell’s reasons for dismissing emotion as a scientifically unsalvageable folk psychological concept might not apply to moods, and hence core affect, too. Is core affect or mood a helpful concept? Does “core affect” cut reality at its joints? To answer these questions, we need to know what makes psychological concepts helpful in the first place. In general, a concept is scientifically helpful if it is systematically embedded in a helpful theory. Many who comment on science (including many scientists) claim that predictive value is the criterion by which theories should be judged. Part 3 of Cognitive Productivity describes several other criteria, however. In any event, core affect is something like the concept of gravity, in that it hangs together with a bunch of other concepts to help make sense of and predict behavior.
Imre Lakatos was the first to systematically argue that to prove its worth, a theory needs to become part of a progressive research programme. Theories are not static. They evolve with a research programme. Lakatos, who focused on physics, assumed that this evolution was based on empirical criticism. However, cognitive scientists believe (or at least ought to know) that in psychology, this means that a good theory needs to be amenable to computer simulation. AI researchers need to simulate the theory, ideally trying not merely to make predictions about behavior but to replicate capabilities the theory touches upon, using the designer stance
ASIDE “design stance” vs. “designer stance.” David Marr described computational psychology in terms of levels of analysis. Daniel Dennett described the design stance. But Marr and Dennett specifications of AI had some flaws that Aaron Sloman addressed. Sloman first used Dennett’s label, the “design stance” to refer to a set of AI methodologies, but later shifted to the “designer stance”, due to the differences with Dennett’s view. Dennett later claimed that his own design stance was the same as Sloman’s “designer stance”. However, John McCarthy (co-founder of AI) agreed with Sloman and used the concept in “The well-designed child”, published in the Journal of Artificial Intelligence (2008). For example, Dennett’s three stances [intentional, design, and physical stances], like much philosophy of science, are focused too narrowly on prediction and insufficiently on understanding and capabilities. The best cure for this, IMO, is actually doing AI on complex problems, including computer programming of symbolic systems.
To develop AI models of core affect, one might ask (in Marr’s terms), What is the core computational problem solved by core affect? What type of information is integrated by the human brain in the psychological construction of core affect? I would expect this task to involve the integration of manifold substates or measures of energy and tension (Thayer’s apparently equivalent model). Let us call these measures (calculated by the brain, potentially modelled by computer programs) “infra core-affect”: inputs to core-affect. For example, there might be a collection of mechanisms, each one of which computes tension in particular locations of one’s body. These tension computers might be organized in (hierarchical?) networks. Another set of computers might integrate various inputs to pain sensations.
In designing a program to compute core affect, we would come to realize that core affect depends on the integration of many inputs. We might then be led to entertain the conjecture that the brain can helpfully integrate subsets of these infra core-affect states.
Perhaps core affect could be conceived as a shifting of infra core-affects. This reminds me of Dr. Murray Shanahan’s Embodiment and the inner life: Cognition and Consciousness in the Space of Possible Minds. It presents a dynamic network theory of the conscious condition. (My critique here.) He argues that consciousness experience is based on competition of (shifting) coalitions of networks. This in turn is based on Bernard Baar’s theory of consciousness.
Following Dennett’s “multiple drafts” theory of consciousness, one might wonder (i.e., skeptically question) how useful it is for the brain to integrate these states. One might even come to think that this supposed integration is illusory. There does not necessarily need to be a global integration of core affect.
This, then, points to another avenue of progress for Russell’s theory of core affect. It needs to be integrated with computational theories of consciousness.
These considerations are relevant to the theory of sleep onset and insomnia that Hyniewska and I presented at ISRE. Most psychological and biological theories of insomnia state (or imply) that “hyperarousal” is at the root of insomnia. This is what I call the “Arousal Dogma of Insomnia”. Our Somnolent Information-Processing (SIP) theory challenges this dogma from the designer stance.
But what does this mean for our original question about moods and self-knowledge? Does one need to read reams of psychological research literature to make sense of one’s moods? No. But I do think it helps to know some helpful theories of moods, and to seek to improve one’s understanding of them over the years. That can help one better understand oneself and one’s loved ones. To this end, I recommend Thayer’s Calm Energy book, Russell’s articles and the other publications mentioned above.