4 March 2008
Philosophy of
Biological and Cognitive Sciences
Matteo Mameli and
David Papineau
Tuesdays 12-1.30 Lecture Room KCL Dept of
Philosophy
KCL/LSE MSc in PHS
Paolo Mantovani
Cultural evolution 2:
Memetics and the problem of cultural replication
Last seminar
i) EC
cannot explain inter-group behavioural variation in
humans.
a)
Inter-group genetic variation underdetermines inter-group behavioural
variation
b) Variation
in the natural environment
underdetermines inter-group behavioural variation
c)
Something else must be the primary explanans (=
cause) of inter-group behavioural variation → cumulative culture most plausible
candidate.
ii)
Selection for cumulative culture
(cumulative culture as an adaptation).
→ How
should we frame processes of cultural transmission? Evolution
or simple change?
Cultural evolution theory
(CET) (Variational model): a) Memetics
b) Co-evolution.
Memetics
- Dawkins (1976) then
Dennett, Blackmore et al.: meme as second replicator once
extended capabilities for social learning/imitation are in place. ‘a unit of cultural inheritance, hypothesized
as analogous to the particulate gene, and as naturally selected by virtue of
its phenotypic consequence on its own survival and replication in the cultural
environment’ (Dawkins).
- Memes (from Greek mimeme = ‘the one that gets copied’): (generally) ideas,
beliefs, skills, behaviours, values etc. stored (for
the most part) in brains/minds.
-
Memes sustain evolutionary
algorithm: variation, inheritance, differential replication.
-
Memetic inheritance: the mean = imitation/social learning.
Cultural parents instead of biological parents.
-
meme-eye-view: cui bono? The memes (individual
psychologies are the vehicles) →
strict analogy with gene-eye-view and genetic evolution but process largely autonomous
from genetic evolution.
-
Ex: religion as a
‘mental virus’ (Dawkins, Dennett). Or
other specific explanations (Blackmore): agriculture,
human brain size etc
General problems for memeticists:
informality, tautological explanations? → no
adaptive explanations (ex. sociobiology) but no autonomous causal framework.
Issues debated within memetics:
-
What is the
cultural analog of the genotype/phenotype distinction? (belief/behaviour?)
-
What is a meme? (neurological
patterns? Behaviours? Beliefs? What about artefacts?)
-
What is the size
of a meme? (3-notes pattern or a symphony?)
-
What kind of social
learning/imitation involved? (motor patterns automatic
imitation or social learning/imitation in the broad sense?)
-
…
One important problem: are memes copied with high-fidelity? Is the
question decisive for the suitability of a selectionist/variational
model of cultural processes? (This is a question that potentially affects all evolutionary approaches to culture)
-
Dawkins’ view: copy-the-instructions
vs copy-the-product
-
most times we copy explicit instructions (the written
recipe of the cake) or take others’ behaviours as
instructions (your mom making the cake) rather than copy direct products (the
cake). Instructions (rough analog of the genotype) are ‘self-normalizing’ and
grant continuity at the level of products (rough analog of the phenotype).
Sperber’s critique of memetics
a)
A variational/selective model requires high-fidelity replication of particulate units
b)
Cultural
transmission (for the most part) does not involve replication of particulate
units
c)
Cultural transmission (for the most part) is not a selective process
Focus on (b): Cultural transmission
is a re-production
process rather than a copying process → we infer others’ mental representations from public
representations (bodily movements, uttered or written words, artefacts etc.).
Simple
catchy memes a part (e.g., a pop tunes), it is very likely that on a 1-1 basis
cultural re-production involves high rates of mutations. Why? No access to others’ mental
representations and the evidence (public representation) underdetermines the
reliability of the product (mental representation).
→
(c): most times selection will not be effective → no
cumulative selection. 1) Mutation
sweeps away selective pressures on representations.
However, how do we explain cultural
stability if there is continuous mutation?
Sperber’s reply: cognitive ‘attractors’ (1996) = ideal representations (types) to which
the real representations of the individuals converge in virtue of shared,
mostly innate biases (e.g. UG, structures in story-telling). (Cf. poverty of
the stimulus argument for UG).
-
Kinship with evoked
culture. But this view is more sophisticated: attractors have strong
genetic base but they can also be ‘socially constituted’ at various levels (even
though Sperber and other cognitive anthropologists
usually focus on the first ones, relating to modules such as ToM, face recognition, folk biology etc.).
-
Reply to Dawkins:
instructions are not properly ‘self-normalizing’,
they have normalizing effects as long as they are interpreted as instructions → attributions of intentions
behind public representations through shared types/attractors. Social
learners share attractors → cultural stability.
The scribble/star example.
-
→ rich imitation (e.g. ToM) → no
proper replication because much information is in-built and not properly
gathered from the model (Sperber → condition for proper replication: ‘the
process that generates B must obtain the information that makes B similar to A
from A’).
-
Side question: A part from cultural
stability, how does the attractor model account for cultural variation
and the dynamics that bring it about?
According to Sperber the attractor model adds another problem for the
selection model.
2) Attractors, again,
sweep away selective pressures on representations.
Dennett’s ‘Thinkos’
Dennett → Sperber uses a narrow definition of replication that
does not apply to genetic transmission itself. DNA chunks use ‘type-knowledge’
to recognize and produce type-like things’ (like copies of themselves and proteins).
There is no brute copying. Rather, DNA uses an alphabet to recognize chemicals
and rules to produce pre-wired outcomes: ‘the high fidelity of genetic
transmission depends on the subcellular machinery
being triggered to ‘recognize’ and ‘re-produce’ a small repertoire of types,
whose idiosyncrasies, if any, are ignored, not slavishly copied’.
-
the type-knowledge involved in cultural transmission is
more complex but not different in kind. ‘Thinkos’
= rules to decode intentions/instructions/meanings from public representations
= higher level ‘typos’.
-
Sperber is right in emphasizing the re-productive feature of cultural transmission but wrong in
thinking that this constitutes a decisive problem for the selective model. He
maintains strong analogy genes-memes, both replicate (but no in Sperber’s naïve sense).
-
However problems for
memetics as a science: the complexity of the ‘thinkos’ and definition of identity conditions for
memes.
The view from co-evolution
Heinrich and Boyd challenge
both Sperber’s conclusions. a) High-fidelity
replication at the individual level is not a necessary condition for the
suitability of the selective model. b) Attractors can facilitate selection. Two models.
1) Assuming continuously varying representations,
attractors facilitate selection. Actually, the stronger the attractors the
more selection is effective. Attractors render the cultural environment more
simplified, it washes away ‘noisy transmission’ . If there is a force that biases
selection of one particular attractor this force will be more effective because
of the simplified environment. If the variants available vary through a continuum
the risk of random choice increases (all other things being equal).
2) Assuming high rates of
mutations of particulate cultural
variants, conformist bias can explain
cultural stability (and the possibility of cumulative retention).
Assume high rates of mutation
(as Sperber suggests) → in a genetic model this
would prevent stability. However we don’t need to push the genetic model too
far in matters of culture! Conformist bias at the individual level has the
effect, at the multi-generational population level, to correct the inaccuracy of
1-1 transmission. The most common cultural variants will still increase in
number over time, other things being equal, and stabilize culture, even
assuming high error rate.
-
Also, effects of explicit
teaching (→ more accurate transmission): ex. selection for explicit
teaching in vertical transmission (Mameli) → it increases inclusive
fitness.
INTRODUCTORY:
Laland N. L. & Brown G. R. (2002), Sense
and Nonsense. Evolutionary perspectives on human behaviour, Oxford
UP, chapters 6,7.
COGNITIVE ANTHROPOLOGY:
Sperber D. (2000), ‘An Objection to the Memetic
Approach to Culture’, in Aunger R. (Ed), Darwinizing Culture,
Sperber D. and Hirschfeld A.
(2004), ‘The Cognitive Foundations of Cultural Stability and Diversity’, in
Trends in Cognitive Sciences, vol. 8 no. 1, pp. 40-46.
Sperber D. and Hirshfeld L. (2007),
‘Modularity and Culture’, in T. Simpson, P. Carruthers,
S. Laurence and S. Stich (eds), The
innate Mind: Culture and Cognition, Oxford UP.
Sperber D. (1996), Explaining
Culture: A Naturalistic Approach, Blackwell,
MEMETICS:
Dawkins R. (1999), ‘Foreword
to Blackmore S., The Meme Machine,
Dennett D. (2000), ‘From Typo
to Thinko: when evolution graduated to semantic
norms’, in Levinson S. C. and Jaisson P. (Eds), Evolution and Culture, MIT Press, pp. 133-145.
Dawkins R. (1976), The Selfish Gene, Oxford University
Press,
Dennett, D. (1996), Darwin’s Dangerous Idea: Evolution and the
Meaning of Life,
Blackmore S. (1999), The Meme Machine,
CO-EVOLUTION:
Henrich J. and Boyd R. (2002), ‘On Modelling
Cognition and Cuture: why cultural evolution does not
require replication of representations’, in Journal of Cognition and Culture,
2.2, pp. 87-112.
Richerson P. and Boyd P. J. (2005), Not by Genes Alone: How Culture Transformed Human Evolution,
Mameli M. (forthcoming),
‘Understanding Culture: A Commentary on Richerson and
Boyd’s Not by Genes Alone’.