Ths file is
http://www.cs.bham.ac.uk/research/cogaff/misc/messages-to-isre.txt

It includes two messages posted to ISRE (International Society for
Research on Emotions) on 17 Jul 2005 in answer to a requst for help from
an emotion researcher.

It immediately generated some favourable responses, so I thought it
might be of wider interest.

I may later produce a version in HTML.

Aaron
http://www.cs.bham.ac.uk/~axs/


From Aaron Sloman Sun Jul 17 12:22:26 BST 2005
To: ISRE-L@LISTS.PSU.EDU
Subject: Re: [ISRE-L] Individuating emotions

Kristján Kristjánsson asked for help relating to individuating 'real'
emotions.

> A feature of cognitive theories of emotions, which many of us
> philosophers interested in emotions find unsettling, is their potential
> for endless, uncontrollable proliferation of kinds of emotion, or - put
> in different terms - their lack of rigor in individuating 'real'
> emotions. It does not matter here whether we understand the cognitive
> element of emotion to constitute a judgment, belief or construal; there
> seems at first glance no limit to the kinds of emotion we can posit by
> combining a judgement, belief or construal with a desire/concern and
> affect.

I think this is a very important, largely unnoticed, point about the
variety of phenomena which people call emotions.

The explanation offered below is that we are dealing with a 'generative'
system, analogous to generative grammars. Apologies for length.

A similar point can be made about motivation, attitudes, preferences,
ideals, and other kinds of affective phenomena, as well as more general
cognitive phenomena such as intentions, beliefs, imaginings, puzzlement,
perceptions, etc.

A failure to understand the reasons for the 'endless proliferation'
leads to attempts to classify emotions (and other things) in terms of

EITHER

    sets of dimensions (e.g. each emotion is a combination of some set
    of attribute values -- often limited to two dimensions because
    that's all people can easily draw in their diagrams or visualise!)

OR

    a taxonomy which is a tree generated by a collection of questions
    about the emotion, where branches at a node of the tree represent
    different answers to the question asked at that node

The taxonomy approach is more powerful because it allows different cases
with different kinds of complexity, whereas the dimensions approach
leads to fixed kinds of complexity: every case is represented by a
vector of features, where all the vectors have the same number of
features.

But BOTH approaches are, in the long run, of limited scientific value
because they (usually) focus on 'surface' features of emotions (etc):
things that can be observed or measured, rather than on the kinds of
*underlying* structures and mechanisms that are capable of generating
the variety under study.

Compare

A: ancient attempts to classify kinds of physical matter, and kinds
    of processes, in terms of what we observe or measure about them,

with

B:  the modern approach of the physical sciences, that postulates an
    ontology containing a variety of structures (e.g. molecules, atoms,
    and subatomic particles) that can exist in many different kinds of
    relationships whose complexity varies depending on the number of
    components (e.g organic vs inorganic molecules), where the variety
    of processes that can occur depends on how the components of
    different structures interact and change their relationships thereby
    consuming or producing ambient energy or matter and producing
    effects of very different types and very different complexity.

In general, explanatory constructs of type B go far beyond what can be
directly measured or observed and depends on creative leaps generating
novel hypotheses that can rarely be directly confirmed or refuted (with
or without the use of statistics!), but which may or may not (as Imre
Lakatos documented) survive a lengthy period of critical analysis,
testing, and refinement in competition with other explanatory
hypotheses.

Sometimes only future technology can yield some of the data needed to
choose between rival explanatory hypotheses if they are really creative.

My impression is that much of the study of emotions (and other
psychological phenomena) still uses conceptual frameworks like A, above,
(often because that is mistakenly taken to be how physics works),
whereas we actually need new analogues of B, where the underlying
components are not necessarily physical and chemical structures and
processes (though at bottom, that's the implementation level), but more
abstract structures and processes, such as occur in complex virtual
machines in information processing systems.

These abstract virtual-machine components can be forms of
representation, semantic contents, mechanisms for acquiring, storing,
transforming, combining, analysing, matching, decomposing, inferring,
and using many different kinds of information in generating,
controlling, modulating, or terminating, many different kinds of
processes -- often several processes in parallel with emergent
interactions between them.

[
Note: the concept of 'information' used here is no more explicitly
definable in terms of more familiar primitives than the concept of
'energy' in physics, for reasons explained in this presentation:
    http://www.cs.bham.ac.uk/research/cogaff/talks/#inf
]

Thus, instead of construing anger as (for example) a mixture of
observable or measurable, physiological, behavioural and/or neural
processes, we can construe different sorts of anger as involving
different kinds of interactions between percepts, motives, values,
beliefs, expectations, desires, intentions, and various kinds of
voluntary and involuntary control processes, occurring simultaneously in
different parts of a complex information-processing architecture.

(A first draft theory can use common sense categorisations of these
underlying mental structures and processes, such as might be given by a
novelist or garden-gate gossip -- but we'll need to progress beyond
those to deeper explanatory constructs -- such as those being explored
in AI, clinical and developmental psychology, linguistics, etc., though
there's still a long way to go.)

Variations in kinds of structure and complexity of states of anger,
fear, hope, joy, dismay, surprise, pride, jealousy, infatuation,
schadenfreude, etc., are closer to the kinds of variation in sorts of
structure and complexity of interactions between complex molecules than
anything that is easily represented as a node in a taxonomy or as a list
of features.

(A small subset is analysed usefully in
    A. Ortony, G.L. Clore, A. Collins,
    The Cognitive Structure of the Emotions,
    Cambridge University Press,
    1988)

The notion of a 'grammar' that is capable of generating an infinite (or
astronomically large) variety of sentences of many different structures
and functions from a relatively small set of components and recursive
relationships, may be a useful analogy to what we need to understand
the variety of emotions. I produced a first draft version of this idea
in this paper in an obscure journal

    Towards a Grammar of Emotions,
    in New Universities Quarterly, 36,3, pp 230-238, 1982.

now available online in postscript and pdf formats here:

    http://www.cs.bham.ac.uk/research/cogaff/0-INDEX96-99.html#47

Thinking about the question 'Why can't a goldfish long for its mother?'
may lead lots of people to similar conclusions.

(DIGRESSION:
I wish everyone would put their papers online so that they are easily
accessible both to search engines and to people who don't subscribe to
journals with a narrowly focused readership. Some ISRE members already
do that: it would be good if ISRE were to recommend all to do it. I
always cross out bits of copyright forms that prevent me doing that and
I have never met a publisher that objected. Some publishers now even
explicitly allow it in copyright forms, e.g. Elsevier, and since most of
the research is paid for from public funds, the research results should
be freely available to all sorts of tax-payers, especially in
developing countries, which are exploited in many ways by richer
countries, e.g. by employing their trained doctors and nurses and
science graduates.).

[Kristján]
> What distinguishes those cognitive elements that can form the
> basis of an emotion from those which cannot, and how can be distinguish
> clearly between different emotions? Why is 'angling indignation' (the
> indignation of an unlucky angler who sees other, less qualified, anglers
> catching fish all around him), for instance, not a specific emotion? Or
> is it?!

In the framework suggested above this question can be seen as analogous
to the question whether stuff composed of a particular type of complex
molecules is or is not a specific kind of chemical.

The answer is clearly 'yes'. Whether it is of any *interest* to give it
a name or not (apart from its structural representation) may depend on
what kinds of causal interactions there are between that chemical, or
emotional state, and other kinds of things. For instance, 'quinine' is a
label given to a substance with a particular moderately complex
structure, depicted here

    http://sres.anu.edu.au/associated/fpt/nwfp/quinine/Quinine.html

because it happens to be quite a useful substance!

If, for some reason, angling indignation started manifesting itself
frequently in violent attacks then newspapers might invent a name
for it, as has already happened with 'road rage'.

I think there is a partly analogous phenomenon that is already well
known but has no name, but could be called 'patient rage' that some
unfortunate doctors have experienced. Similarly 'claimant rage' in
social security offices.

[Kristján]
> Notice that I am not interested in the question of which emotions are
> most 'frequent' or 'important' or psychologically 'basic', but rather
> the deeper epistemological (or even ontological) question of how to
> individuate emotions.

I suggest that the ontological question is best addressed in the
framework of a study of architectures and mechanisms of mind and the
kinds of states and processes that can be generated within such systems
(including processes with varying aetiology and time-scales).

That will indeed reveal an infinite (or astronomically large) variety of
cases.

Which special cases are worth labelling and studying will, as in
the case of chemical substances and processes, depend on how they
interact with other things we are interested in.

I have not yet fully digested Tom Scheff's paper mentioned in this
thread
    http://www.soc.ucsb.edu/faculty/scheff/47.html
    A Taxonomy of Emotions: How Do We Begin?
    Thomas J. Scheff

but I think some of his ideas are closely related to those above.

However we have a disagreement on this point in his paper:

[T.J.S]
> In particular, for the development of a concept, how many parts as
> compared to how many wholes? It doesn't seem likely that there should
> be many more wholes than parts, as in Lazarus's chapter 8. More
> likely, there should be many more parts than wholes, ....

A generative system in which components can be combined in various
relationships to form larger wholes which in turn can be combined with
other larger wholes (as in chemical compounds) will produce far more
wholes than parts. This is the source of Kristján's question!


If these points were generally accepted, the sciences of mind might be
transformed and integrated into something deeper and richer than any of
the current strands, and many pointless disputes resolved.

I am not optimistic --  having tried to communicate ideas like this for
about 30 years with very little success.

Aaron
PS
I'll post an example from round the world yacht racing separately.
http://www.cs.bham.ac.uk/~axs/

From Aaron Sloman Sun Jul 17 15:16:36 BST 2005
To: ISRE-L@LISTS.PSU.EDU
Subject: Re: [ISRE-L] Individuating emotions (promised example)

In responding to Kristján Kristjánsson's query about

    "endless, uncontrollable proliferation of kinds of emotion"

I wrote:

> I'll post an example from round the world yacht racing separately.
> http://www.cs.bham.ac.uk/~axs/

Here goes:

Often uncreative interviewers and broadcasters ask people who have been
involved in newsworthy events inane questions whose answer is totally
obvious, like 'How do you feel?' said to an olympic winner.

Occasionally the answer produces something interesting.

Some years ago I listened to a radio broadcast at the end of a
round-the-world yacht race. An all-women team had just finished the
race, coming close to winning, and the captain whose name I forget
(perhaps it was Clare Francis) was asked how she felt, as she came
ashore.

I did not record her answer, but it had a long list of components such
as these (though I don't recall the order):

    very happy to have finished and done so well

    very disappointed not to have won the race

    very hungry because food ran out a few days ago

    looking forward to a good meal awaiting her

    very pleased to be about to see friends and family again

    very sad that members of the close-knit team on the boat would no
       longer be together

    looking forward to taking part in another planned race
        (my memory of this one is less certain than the others)

she could have added:

    aware of having a very complex mixture of feelings

but I don't think she did!

She did not say she was cycling through those states, having one of them
at a time (as would be required by some emotion theories).

Nor did she say she had a blend of the sort obtained by mixing colours,
where none of the original colours remains, as in vector summation (as
would be required by some other emotion theories).

I have no reason to believe she was lying, or inventing her answer for
the sake of the interviewer.


Perceptive novelists often give their characters states of mind in which
complex, enduring, processes interact strongly (e.g. love for a sister,
hurt by a lover's unfaithfulness with the sister, anger at the sister,
desiring revenge, guilt at wanting revenge because of religious beliefs,
etc. etc.). Novelists do this not as some kind of science fiction, but
because they observe that human minds allow such complex interacting
states and processes. These things are not discovered in laboratory
experiments designed for collecting numbers to feed into statistical
packages.

A good theory of mind needs to explain how this is possible.

The yachting example, and the (invented but plausible) sister example,
are more complex than the 'angler indignation' mentioned by Kristján but
I think they make a closely related point: human minds use mechanisms
that permit the coexistence of multiple, enduring, partially
interacting, states of many kinds, of which we are mostly unconscious
normally (e.g. most people are mostly unconscious of the low-level
processes involved in visual perception, speech understanding, language
generation -- e.g. processes involving phonemic, syntactic and semantic
structures).

However we can be aware of some of what's going on.

One of the features of a normal adult human mind is that (unlike very
young children) our information processing architecture has a sub-system
with meta-semantic competence combined with self-observation,
self-categorisation, and self-evaluation mechanisms. So we have a
partial ability to characterise and reflect on our own states of mind.

This is a product of evolution (modified by the self-descriptive
categories absorbed from the current culture), and works well for many
everyday purposes (e.g. asking someone who has been ill if she feels
better, or an oculist asking a patient whether a line looks sharper
through a new lens).

But this self-observation mechanism (or mechanisms) sometimes produces
excessive self-confidence in relation to the scientific or philosophical
study of mind, because we don't realise how limited our self-observation
is and especially how bad we are at retrospectively summarising mental
states and processes.

This leads to a plethora of mutually inconsistent theories of emotions,
consciousness, learning, etc., which are often taken to be 'obviously
true' when in fact they are false, like the often-made claim that if you
have an emotion then you are aware of having it. (Novelists know better
than that.)

Of course, it's not just emotion theorists. Immanuel Kant, who inspired
much of my thinking about minds, wrote that every judgement has two
components, a subject and a predicate, in a book that was full of his
own judgements with far more complex structures, including conjunctions,
disjunctions, negations, conditionals, nested qualifications, etc.
Accurate self-observation is not easy, especially when one has a theory
to propound. I am sure I am equally guilty.

Cheers.

Aaron
http://www.cs.bham.ac.uk/~axs/

======================================================

NOTE:
In response to the above, David Konstan, in the Classics Department at
Brown University, posted a note responding to my remark that

> Perceptive novelists often give their characters states of mind in which
> complex, enduring, processes interact strongly"

He commented that the practice goes back at least 2000 years, giving
examples.

Also relevant:
    http://www.cs.bham.ac.uk/research/cogaff/misc/isre-levels.html