Today I will summarise the account of pain and pleasure provided in The Emotional Mind. This builds on the account of valent representation that I outlined yesterday. However, the first thing to note that is although valent representation is representation in a valent (i.e. positive or negative) manner, it is not a representation of valence. Indeed, one of the reasons that pain has become such a hot topic lately is due to the recognition that the intrinsically unpleasant aspects of pain are distinct from the representation of damage to the body. This dissociation is mostly justified by pain asymbolia, which is condition in which individuals seem to be able to feel pain, and yet not experience this feeling as bad. The very possibility of this condition forces the distinction upon us. Thus philosophers have been trying to give accounts of what exactly makes an experience intrinsically unpleasant, where some also propose symmetrical accounts for the intrinsically pleasurable aspects of experience. Following Murat Aydede (2014) I use the term ‘affect’ to cover both pleasant and unpleasant aspects of experience. We can then define an affective state as any mental state that involves the arousal of affect. I encourage others to adopt this terminology.
An initial point to make about affect is that there must be some connection with valent treatment. Painful and pleasurable affect must involve approach and avoidance responses respectively (even if we’re just talking about cognitive kinds of response like making plans or recalling details). If they did not, we would have no way to distinguish affect from non-affective sensations, and no way to oppose pain with pleasure. This is not to deny the existence of cases where we might embrace a pain or avoid a pleasure, but those should be understood as complex and derivative cases. Thus valent treatment is necessary for affect. However, this allows for two options: one where affect is involved in directing the valent treatment, and another where affect is involved in tracking the valent treatment. These options correspond to two broad approaches to affect: evaluativism- where affect describes the condition of the body, and imperativism- where affect directs the subject to do something (see Colin Klein for a prominent imperativist view)
My approach is basically evaluativist. There are various reasons for this, but one of the major ones is that imperativism does not do well with pleasant affect. This is because a mental state in which one strongly or urgently directs the subject to deliver a certain good is compatible with lacking the good (or the experience of the good) entirely, and such states of yearning are typically quite unpleasant. What we also need for pleasant affect is for such a motivation to co-exist with actually having some of the good thing. Thus, pleasurable affect is essentially a representation of success; that the organism is directed towards acquiring some good, and has managed to do so. I call this a success theory of pleasure. I then offer a symmetrical ‘failure theory’ of pain, where the subject is directed towards avoiding some harm, and yet the harm is still present.
The upshot is that affect is a higher-level representation that combines a representation of harm or benefit, with a representation of the directive response. Thus the model is that on top of regular valent representations (see Tuesday’s post) there’s a higher level node that monitors the presence of the good and the response. This is illustrated below.
Meanwhile, affect itself follows the structure of a negative feedback loop because it also automatically triggers a response of increased attention.** Again, there are various reasons for thinking that affect directly stimulates attention rather than specific responses, but the main one is that organisms already need to automatically respond to harms and benefits prior to developing the capacities for pain or pleasure.
Anyway, since affect automatically triggers attention, it is an intrinsically motivational state, but because increased attention is common to both pain and pleasure, this response cannot distinguish the two. Rather increased attention indirectly serves a valent function by boosting or prioritising the underlying aversive and appetitive responses.
A final aspect of the account that I am pleased with is the following hypothesized equation for affect intensity:
Affect = response x object
That is, the intensity of pleasurable or painful affect is equal to the (detected) intensity of the response multiplied by the (detected) degree to which the harm or benefit is present. This means that if either the object or the response level is detected as 0, the subject should experience no affect. I think this equation could be a significant psychological principle, particularly given the role of affect in directing attention, so I am hopeful that it will be empirically confirmed.
** A claim not in the book, but which I’ve been strongly considering lately, is that affect is the only driver of attention.
References
Aydede, M. (2014). How to Unify Theories of Sensory Pleasure: An Adverbialist Proposal. Review of Philosophy and Psychology, 5(1), 119-133.
Big problem with the pain/pleasure divide – is it still universally used? It’s simplistic. There are actually TWO families of positive feelings, and they’re in more or less continuous conflict, and this is written into the design of the body. (Haven’t thought whether pain also needs dividing). There are feelings of SATISFACTION and there are those of PLEASURE – the feelings of activity vs passivity – high-intensity/low sensitivity activity vs low intensity/high sensitivity activity. Basically we’re in continuous conflict between being active – forms of physical/mental exercise – and being passive – relaxing and consuming physically (food, drink, drugs etc) and mentally (watching tv, reading etc). This is written into the continuously conflicted sympathetic vs parasympathetic nervous systems and by extension, the conflicted upper body vs lower body which they respectively stimulate. Or, as Shakespeare puts it, “But to the girdle do the gods inherit; Beneath is all the fiend’s”. Activity typically brings immediate discomfort and even pain as we adjust to the demands of muscles warming up, or pushing beyond immediate limits. But then it yields positives in the form of excitement and feelings of aliveness and all those eventual endorphins that come with accomplishment and an exhausted body. Passivity brings the immediate feelings of relaxation, and various pleasures of consumption, but it is liable to lead eventually to feelings of satiation, emptiness, and the guilt of non-achievement and muscles unused. Activity and passivity bring radically different and incompatible kinds of positive feeling as per the conflicting actions and rewards of the sympathetic vs parasympathetic nervous system and active vs relaxed muscles. You have to distinguish between *satisfaction* and *pleasure.*
Again, as per our last exchange – but pls comment, because I don’t know the field well – I would suspect that analytical philosophy has enormous difficulties with the idea of deep-seated conflict, because conflict simply doesn’t fit into existing cognitivist, mechanist paradigms. There are no machines or computers with conflicted anatomies. But I suspect if we ever get around to seriously trying to design autonomous robots that can truly live by and for themselves, we might well want to embed conflict in their design as it is in that of living creatures.
P.S. The physiologically perfect activity is sex. – in which we get to experience both the satisfaction of intense exercise and the pleasure of intense sensation. That’s because the innervation of the genitals is massive relative to any other area of the body. In most forms of exercise, sensation and sensitivity are automatically reduced.
A key point about having an distinct account of affect is to find a way to unify the enormously varied class of pleasures and pains. What could drinking a cool beer, watching tv, enjoying sex, reading something interesting etc. all have in common? And the same for all the things we find intrinsically unpleasant? With an account of affect we can allow that all these things vary in sensation and cognitive content, but there’s a common element.
I generally take there to be 4-5 major sources of intrinsic value: sensory pleasure, power, social attachment/love, relief from pain, and maybe excitement (this last might fold into one or more of the others). Each of these needs a separate account. Nevertheless, I think pleasure, power, and attachment at least all share the basic structure of pleasant affect, alongside their varying features. A book I’m working on now is intended to develop this a bit more, with particular reference to aesthetics.
One further point about conflict. One of the key points I make about affect in the book is that it directs attention. And the purpose of attention is precisely to prioritise certain possible actions, or tending to certain concerns over others (see e.g. Wayne Wu on this). That is, attention is a system for managing conflicts between the various regulatory systems that are simultaneously clamouring for satisfaction. It makes sense that affect directs this because affect occurs when something is going particularly wrong or right.
Hi Tom,
Great posts! My interest was piqued by your suggestion that affect regulates response indirectly via attention rather than directly. Could you say a little more here, whether your reasons for making that claim or the broader significance you attach to it?
Aaron
Thanks Aaron!
Pain and pleasure certainly stimulate attention, so that’s one reason. Another reason is that so many of our responses to pleasurable and painful stimuli are automatic (e.g. withdrawal, wincing, rubbing the affected area for pains, licking lips, reorienting for pleasures), so we may not need affect to do anything but reinforce what’s already there.
At the same time, affect can noticeably motivate person-level intelligent actions. A link to attention can make sense of this. Attention prioritises (intensifies activity in) the processing of the sensation. This allows a representation to cross the threshold from unconscious to conscious. Once the stimuli is conscious, the whole person is then recruited to managing the issue.
Note that when imperativists characterise what painful or pleasurable affect motivates, usually it’s something like ‘more of this!’ or ‘less of this!’. That is, it’s a generic command that needs to be directed into a specific response. By directly stimulating attention, affect motivates us to do *something* urgently. But it’s up to either the conscious person, or the automatic systems to figure out what that something is.
I read with great interest your blog on pain and pleasure. It seems to me that one key aspect is missing in your summary, that is, the role played by dopamine in affects with anticipated positive or negative outcomes, and their impact on triggering attention. Studies show that dopamine, released to account for prediction errors, end up irradiating in the orbitofrontal cortical and basal ganglia areas and strengthening the synapses located in this region. After several such exposures to dopamine, these strengthened synapses encode the expected subjective value of the inputs, responsible for positive or negative prediction errors. However, it seems that dopamine isn’t involved in anticipated punishing outcomes, like pain. Pain affects most likely do not follow the same internal processes as pleasure, though it is still not clear what are the neurotransmitters and mechanisms underlying such an aversive loop. (1 and 2)
I am uncomfortable with your claim that affects are the only driver of attention. There are some loops associated with attention, that don’t necessarily involve affects. For instance, for habitual behaviors in which we only need to evaluate one action, a ‘stimulus-response’ model has been associated with the sensorimotor loop, located in the sensorimotor cortices and involving the putamen. More importantly, many cognitive models (e.g., the free energy theory by Karl Friston or Rescorla-Wagner and their Temporal Difference Learning model) and neurosciences research indicate that attention is triggered mostly by the variance between predictions and reality. Friston suggests that within such a hierarchical brain model, attention is the process that optimizes precision during hierarchical inference, rather than the switching on of the various sensing channels, explaining what triggered this attention in the first place. Said differently, the brain is attracted by inputs and actions which corroborates or refutes someone’s beliefs, rather than by the nature of the sensory input, be it an affect or not. (3)
In this blog, you propose an equation, which describes the intensity of pleasurable or painful affects as equal to the (detected) intensity of the response multiplied by the (detected) degree to which the harm or benefit is present. In fact, several studies have tried to quantify affects by using equations. For instance, a computation model of acute pain was developed in 2004 to measure the impact of the substantia gelatinosa located within the spinal cord. (4) This substance acts as a gate control mechanism, influencing the flow of information to the brain and thus the impact on the pain experience. I can also refer to a study that uses a computational model in which happiness reports are construed as an emotional reactivity to recent rewards and expectations. (5) This study shows that momentary happiness is related to quantities of dopamine release, associated with temporal difference errors, signaling changes in the long-term expected reward, and is expressed by an equation.*
If I compare the equation on happiness to yours, we can see that such pleasurable affect is indeed function of the response, which can actually be measured through the levels of dopamine in the striatum. (6) Both equations also account for a detected degree of expected rewards (though the happiness equation doesn’t use multiplication). However, in my view, your equation doesn’t integrate an important factor: the variation between the expected and detected rewards. In fact, what is described as a temporal difference error in the happiness equation, accounts for the dynamic assessment process of the benefits and harms. Neuroscientific researches confirm that this prediction error variance is encoded as voltage modulation at the synaptic level.
Serge (7)
* Happiness (t) = w0 +w1 sum y (t-j) CRj+ w2 sum y (t-j) EVj + w3 sum y (t-j) RPEj
Where the sum is from 1 to t, CR means Certain Rewards, EV Expected Values of certain gambles and RPEs the difference between experienced and predicted rewards, t is the trial number, W0 a constant and other weights W captures the influence of different event types. I apologize, the comment format doesn’t take mathematical equations.
1. Tobler, A. Dickinson and W. Schultz, “Coding of Predicted Reward Omission by Dopamine Neurons in a Conditioned Inhibition Paradigm,” Journal of Neuroscience, vol. 23(32):10402–10410, 12 11 2003.
2. M. Ungless, “Dopamine: the salient issue,” Trends Neurosci., Vols. 27(12):702-6., 12 2004.
3. K. Friston, “The free-energy principle: a unified brain theory?,” 13 01 2010.
4. Karen Prince, Jackie Campbell, Phil Picton, Scott Turner. “A computational model of acute pain”;
5. R. B. Rutledge, N. Skandali, P. Dayan and R. J. Dolan, “A computational and neural model of momentary subjective well-being”.
6. Y. Niv, Reinforcement learning in the brain, 2009.
7. S. Van Themsche, “What’s on their mind? Biological and artificial intelligence”; 2018.
Hi Serge. Thanks a lot for your very informed comment! You will note that my account is pitched at the psychological rather than neurological level of explanation, and different neurological stories could fit the same psychological model. For instance, it’s perfectly possible that different neuro-transmitters are involved in the affective aspects of pain and pleasure, since this still allows that there is a functional symmetry. At the same time, the processing asymmetries between pleasure and pain may be entirely attributed to the *sensory aspects* of these processes, rather than their affect. For instance, we know that pain has dedicated sensory channels in a way that pleasure does not.
I want to emphasise here that motivation is not the same as the intrinsically pleasant aspects of an experience. It is possible that dopamine is related to either, or both, but we need to be a bit careful here. I’m certainly not comfortable with equating the quantity of dopamine release with the level of pleasant affect.
With regards to the claim that affect drives attention, actually, I am including ideas about predictive coding when I make my claim. We can understand the attention we give to novelty (or prediction errors) as a case where the failure to predict arouses negative affect, which then drives attention. I have a section in Chapter Four of my book where I talk about the emotion of surprise and predictive coding.
It is interesting that you link attention with habitual behaviours. I’m not quite sure what you mean by this however- perhaps you could supply an example?
With regards to my intensity equation- it is of course very simple. No doubt the true picture will be more complex. The main point is to capture the proportionality between affect, response and detection. However in the equation you offer, I am unsure if ‘momentary happiness’ can be fairly identified with pleasurable affect. I would have to dig into that paper to evaluate that.
Finally, the difference between expected and detected rewards. Here again, we must be careful about whether dopamine correlates with motivation or pleasant affect or both. However, I do allow in the book that the anticipated presence of a good/bad can arouse affect (e.g. transient pain). In my view, this anticipation would plug into affect in the same way as the detected presence of the good/bad. However, there may well be comparisons between expectation and reality that modulate this input in a dynamic way, so I’m open to modifying the details on this.