TITLE: posting to ISRE-L-at-LISTS-dot-PSU-dot-EDU
Subject: Re: [ISRE-L] Positive negative emotions
Aaron Sloman
Last updated: 21 Jun 2006

SUMMARY

Human motivation is a product of very complex mechanisms interacting with very complex physical and social environments.

If we find certain sorts of motives (e.g. wanting to watch films that make you sad, or wanting to do things that are very dangerous) puzzling then we need to understand the mechanisms the produce and change human motives instead of simply explaining them as if they were products of rational decision making.

In particular there are clearly very powerful motives that can be extremely harmful to the individuals concerned for instance, addictions of various sorts, including chemical addictions and addictions to things like gambling.

My suggestion is that the mechanisms that produce addictions may be closely related to, or even the same as, the mechanisms that produce many other kinds of emotions, except that addictions are an extreme case of their operation.

To understand all this we'll need to understand the information-processing architecture produced by millions of years of biological evolution. Introspection, and the traditional methods of philosophy (conceptual analysis) and psychology (looking for behavioural correlations) are unlikely to produce deep theories of what the architecture is. For that we need to adopt the design-based approach.


My comments below were provoked by these postings to the ISRE list on the topic of 'positive negative emotions'.
AV wrote:
Tuesday, June 20, 2006 7:36 AM
> Who can help me to answer the question why people like to hear
> sad music or why (especially women?) like to watch tearjerkers?
R.B. wrote, the same day:
> It is an interesting and often overlooked fact that people pay good
> money to experience "negative" emotions like sadness, fear, anger,
> even disgust.
Notice that, as someone else remarked, that is true even of uncontrollable terror and real danger -- on roller coasters, etc.

Warning: I am quoting selectively to save space. Join ISRE to see everything.

[R.B.]
> I think of this in term of emotional education: that we are
> motivated by curiosity to explore, learn about, and become
> competent in the inner environment of feelings and desires just
> like the external environment. In this view, there are no
> "negative" emotions, in that the feelings and desires per se are
> not unpleasant. Instead, emotions are informative, and it is when
> they inform us of bad things that they are negative.
Other contributors have offered their own explanations in terms of what people experience, why they choose things, reasons, justifications, etc.
[T.S.]
> One problem, having made these distinction:  it seems necessary
> to explain in some detail what we mean by each of the four terms.
> That is, to develop CONCEPTS of emotions instead of depending
> entirely on the vernacular emotion names.
and other vernacular names, like 'intend', 'believe', 'attend', 'prefer', 'deliberate', 'decide', 'regret', 'notice', 'learn'... and many more.


There is another way to proceed, which is to introduce a sub-personal theory-based ontology to make sense of, fill gaps in, remove (some) inconsistencies in, ordinary language, as the atomic theory of matter did for pre-theoretical kinds of stuff ('air', 'earth', 'copper', 'water', 'fire', etc.) It is usually impossible to remove all inconsistencies in a deep theory -- that's a source of progress, as Popper and Lakatos and others have pointed out:

This will make some people think: 'go away you materialist reductionist'. But I don't believe virtual machine concepts of the kinds we need are reducible to physical concepts: though they may be in some sense explicable in terms of deeper, more general virtual machine concepts.
(Some of you have heard this from me before.)

The 'design stance':
Sometimes it is useful not to think of an intelligent individual concerned as the beneficiary or otherwise of motives, decisions, preferences, etc., but something more abstract: e.g. some of the genes involved.

In such cases we need to seek explanations that don't simply adopt the 'common sense' view of the person as doing something for reasons (the 'intentional stance'), but rather treat a person as a complex information processing system with many subsystems doing different things, often with side-effects none of them was specifically aiming at.
(E.g. some illusions and emotional bugs.)

Genes that produce mechanisms that produce motives that increase the likelihood of propagation of those genes may have a tendency to spread even if each individual with that gene would have done better (e.g. lived longer, required less exertion) without that motive.

One of the obvious cases is the set of mechanisms producing desires related to reproduction: successful mating and subsequent care of offspring has great benefit for genes that influence the development of mechanisms creating those desires, but also has a high biological cost for individuals with those mechanisms, for mothers and in many cases fathers also (e.g. birds where male hand female share the feeding).

[Some women say 'never again' after the suffering of giving birth: yet two or three years later their deep biological mechanisms get working to override that decision, even if they remember it well. I have seen that happen very close to home.]

Mechanisms have evolved to drive them to take the actions necessary for reproduction despite the high cost to themselves.

Of course, they can sometimes be countered by other mechanisms, by cultures, by mechanisms that grow under the influence of cultures, etc. It's the most complex multi-functional information-based control system on the planet. (So far.)

The mechanisms that produce motives, pleasures, and pains, in humans are very complex, are of many kinds, and they develop in the lifetime of individuals under the influence of many factors, and can sometimes change in ways that are not only harmful to the individuals but also reduce the reproductive success of the very genes that contribute to producing the mechanisms.

Examples are mechanisms that produce dangerous addictions involving very powerful desires. Those mechanisms were not selected for that purpose: they are too destructive. This must be a side effect of their main functions, or the main functions of other things that interact with them. (Any engineer will know of many examples like that.)

Some of those mechanisms involve chemical interactions, while others have to be explained in terms of the operation of virtual machines (e.g. addiction to gambling?)

This is just a special case of the general principle that any complex design with multiple control mechanisms processing many different kinds of information is liable in some circumstances to produce side-effects that differ from and even undermine the effects the mechanisms were originally selected for: the more complex a software system the more likely it is to have bugs -- 'undesirable' side-effects of some of the interactions of components that in themselves are normally useful.

[That's one reason why governments should never plan to introduce very large and complex systems designed and implemented as a whole -- like the UK's planned system for the health service. The chances of getting such things right, or even finishing them on time and within budget are infinitesimal (like re-engineering the mouse genome to produce cats) no matter how much prior consultation and analysis is involved, and no matter how clever and experienced the designers, simply because of the combinatorics.

The more different sorts of parts there are the more subsets need to be checked: an exponential growth.]

Maybe, just maybe, love of music, of thrills, of tragedies, of suffering of others, of doing mathematics, of dancing, singing, listening to music, watching plays, etc. may be more akin to addictions of various kinds than to rational processes to be explained by reasons.

For me, reading novels has a disruptive effect similar to an addiction, so I don't start any except on holiday. (If I am not too far behind with other things.)

Fortunately, my addiction to some kinds of music (e.g. Bach contrapuntal instrumental music, and Mozart piano concertos) does not seem to interfere with writing messages like this. (Most of the time.) But I don't listen to it because it produces some other effect, as far as I know.

If I did that the same question would arise about why I seek that effect -- leading to an infinite regress of explanations (as many philosophers have pointed out, e.g. G.Ryle The concept of Mind 1949).

> We have done studies with music videos and found that people like
> some videos if the music makes them sad, others if music makes
> them happy, others if music makes them feel powerful, etc. (Buck,
> 1988). But, across all the videos, people liked them if the music
> made them interested and disliked them if the music made them
> bored.
Most of the things you like, you like for their own sake, not because of side effects. Why do I say that? Because you would not want them replaced with something else that produced the same effects. Think about what 'being totally absorbed in solving a problem' (or listening to a fugue) means. If you did it for the side effects, you would not be totally absorbed by the problem or the music. But humans (including my wife) can be totally absorbed in some task (G.Ryle).

Research methodology
Moreover, asking people questions about these postulated benefits of listening to music or doing mathematics things is open to familiar problems: What people are conscious of and can tell you about what they like or dislike is a function of the extent to which they have architectures that are capable of accurate self-monitoring of internal processes. People vary in this. Some kinds of self-monitoring come from intense training as any experienced musician or programmer knows.

We do have some limited common functionality of that sort (e.g. you can answer the optician's questions about which lens makes a pattern look sharper, and record some of what you were recently thinking). A painter who tries to paint how things are as opposed to how they look will produce a six-year old's paintings. Or may be a cubist.

But there's always far more going on than we can report on.

I've watched a four and a half year old child try to copy an outline square without success. He knew he was failing and got upset, but did not notice that his problem was at least partly due to trying to draw all four sides in one movement. If I encouraged him to stop after each side and think where the next one has to go he managed. But he still did not notice what was making him successful in those cases, and got very frustrated: By then he wanted to draw the square, not to get some happiness from drawing it.

Limits to self-monitoring
E.g. most people have no access to the grammatical and phonological rules they use every day in producing and understanding language (e.g. rules about tag questions being negated if the main clause is not negated and vice versa, as in

The train has arrived, hasn't it?
The train hasn't arrived, has it?
[How many people know they use that rule? There are special circumstances where the rule is broken.])

Likewise the person who experiences the Ebbinghaus illusion (seeing a disc as bigger or smaller than others of the same physical size) at the same time as accurately opening finger and thumb exactly far enough to grasp the disc will not know why it looks the wrong size even though the brain clearly has and uses information about the actual size.
[Many attempts to explain this miss the point that visual servoing -- e.g. controlling continuous actions -- needs to be done by different mechanisms from producing re-usable descriptions of the current situation. More needs to be said about why the latter task should generally not use precise metrical information required for the former -- it's partly to do with explosive search-spaces in a continuum.]

So there's no reason to think that asking people why they like something or why they do something will always produce answers to scientific questions: even if the answers are consistent with one another.

(Many people understandably thought the earth was flat once: after all, it looks pretty flat most of the time -- how things look, inside or outside is just more evidence to be explained.)

[R.B.]
> That suggests that listening to music is a kind of exploratory
> behavior where one is exploring one's own feelings and desires.
> The more fear elicited by a horror show, the more anger and sense
> of power elicited by an action-adventure, the more tears jerked
> by a tearjerker, the better they are generally regarded. That is
> the case with much of the phenomenon we term "entertainment," I
> think.
Or maybe things are going on whose explanation is no more amenable to introspective reflection of that sort than the details of our linguistic processing or our visual processing, or our posture control are.

Music has been very important to me most of my life -- both as a listener and as a player, but I don't think I have any idea why, and nothing I've read about it, including that comment, seems consistent with everything I know about music.

Likewise what I've read about emotions. (Partly because there are so many theories that contradict one another.)

Perhaps we need to try build theories about deep mechanisms, where most of the evidence and tests are far more indirect than asking people questions. I.e. we may need to use the 'design stance' in order to get a deep understanding of all of linguistic processing, visual processing and emotional processing (and many other things, like changes in tastes, conceptual development, the growth of self-awareness) even if introspective reports provide some of the data.

Having said all that, I am not yet impressed by most of the current computational attempts to model emotional phenomena that have so far come out of AI: they typically focus on a tiny subset of relevant facts to be explained, e.g. shallow behavioural manifestations, ignoring all the stuff novelists and playwrights know about.

They are not based on deep conceptual analysis, cross species surveys, functional requirements for architectures, and other constraints.

Sorry this is all sounds negative. I think we can and will make progress, but it will be slow and requires kinds of cross-disciplinary co-operation most people don't have time for, because of pressures of funding, tenure, etc. I was extremely lucky not to have those pressures for a long time after I became a lecturer in 1962. As a result wonderful people of many kinds taught me many things. I wish I were better at passing it on.

If anyone has read this far, I apologise for length.

Aaron


Maintained by Aaron Sloman
School of Computer Science
The University of Birmingham