Artificial Intelligence

Author(s): Hugo F. Reading

Source: Man, New Series, Vol. 18, No. 1 (Mar., 1983), pp. 183-185
Published by: Royal Anthropological Institute of Great Britain and Ireland
Stable URL: https://www.jstor.org/stable/2801770
Accessed: 27-06-2019 08:40 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Royal Anthropological Institute of Great Britain and Ireland is collaborating with JSTOR
to digitize, preserve and extend access to Man

This content downloaded from 103.231.241.102 on Thu, 27 Jun 2019 08:40:42 UTC
All use subject to https://about.jstor.org/terms
 CORRESPONDENCE

Artificial intelligence be complex. A complex sufficient condition

Polythetic classification, expounded and discus- takes the logical form 'both p and q must be
sed in Man by Needham (I975), Solheim (I976) present for r'. A complex necessary condition
and Ellen (I976), has a heuristic value in that, if takes the form 'either q or r must be present for
viewed in conjunction with research in artificial p'. Although it is instantiations of generic phe-
intelligence (AI), it helps create an awareness of nomena which are observed, it is connexions
the possibility of the automatic discovery of between generic phenomena which are sought.
causal and nomic connexions between generic The nomothetic method is usually thought of
social phenomena. Such automatic discovery as involving abstraction in its first contact with
goes beyond the computer simulation, automa- objects, in that it separates a partial aspect or
tic classification, cluster analysis and multivari- property from a total object. However, adv-
ate analysis well-established in social science. ances in the description, perception, recogni-
It has been suggested that human survival tion, matching, analysis and manipulation of
would have been impossible without the de- patterns, configurations and structures suggest
velopment of the concepts and classifications that holistic comparisons could serve a
necessary for parsimonious accounts of nature nomothetic purpose.
(Raven et al. I97I). Temkin (i965) speaks of the Popper (I960: 77) distinguishes totalities
universal urge to sort particulars. Piaget (I976) from Gestalt wholes. He maintains that a total-
describes how the child makes dichotomous ity cannot be scientifically studied because any
distinctions and carries out 'grouping' opera- piece of the world is describable by an infinite
tions. If concepts were to characterise entities number of propositions. Thus, it is not a social
richly, the very purpose of conceptualisation totality which needs to be represented. Given
would be defeated. that it is possible to construct different overlap-
Initially, we might suppose that it is ping sequences of connexions and that a totality
monotheticity which results in our perceiving may be selectively and systematically described
and describing the world in terms of concepts in different ways, what needs to be represented
based on only a few characteristics. It is clear, is the resulting configuration. According to
however, that although polythetic classes are Lehnert (I979), it is not yet possible adequately
usually more richly characterised than are to describe a physical configuration, as the
monothetic ones, irrespective of the way in necessary representational system for spatial re-
which characteristics are distributed amongst lations has not been developed. Likewise, it
members, this is not true of the members of the would seem that the adequate description of a
classes. However, the ability of computers to social configuration awaits the development of
handle rich characterisation, as shown by the the necessary representational system for the
computerisation of numerical taxonomy, sug- abstract relations involved.
gests further possibilities where the processing The mathematical theory of patterns,
of complex social patterns is concerned. pioneered by Grenandeir (I976), the study of the
Let us explicate 'causal and nomic connexions properties of systems in general and the iden-
between generic social phenomena'. Included tification of system-forming relations and con-
under 'phenomena' are characteristics (charac- nexions (von Bertalanffy I969; Blauberg et al.
ters, properties, qualities, features, attributes, I977), as well as AI research on knowledge rep-
traits), events, states of affairs, system-states. resentation, should aid the development of such
By 'nomic' is meant any invariable concomi- a system.
tance, a concomitance which need not support The three main statistical schools in
either a counterfactual or a subjunctive con- sociocultural anthropology, the Kulturkreis,
ditional. The connexion may take the form of an California and Yale schools, in spite of a vast
invariable concomitance of properties and so be output of correlation analysis and even some
nonsequential. The connexion may have a multivariate analysis of cross-cultural data, have
probablistic character. failed to develop a methodology for structural
Causation is best viewed in terms of condi- comparisons. Levi-Strauss (I972: 95-6) has
tionship (von Wright I97I; 38 sqq.). Conditions pointed the way by suggesting holistic compari-
may be necessary or sufficient or both and may sons of societies in terms of 'the differential

This content downloaded from 103.231.241.102 on Thu, 27 Jun 2019 08:40:42 UTC
All use subject to https://about.jstor.org/terms
 I84 CORRESPONDENCE

features characteristic of the systems of rela- not limited to re-discovery is the inductive in-
tionships which define them'. It is possible to ference system Meta-DENDRAL, of the family
study the permutations and transformations of programmes known collectively as DE-
amongst structural elements. Intra- and cross- NDRAL (Buchanan I979), which has suc-
cultural comparisons of structures at different ceeded in discovering new chemical rules. The
levels may be made. If we accept the view of rules take the form 'When situation S occurs,
Levi-Strauss that societies consist of diverse, action A occurs'. The programme successively
partially autonomous orders of interrelated en- summarises the chemical input data (IN-
tities, it follows that no society can be repre- TSUM), generates plausible rules through a
sented by a single configuration. heuristic search of the legal rule space
It may be instructive to look at attempts at (RULEGEN), and then refines the rules
automatic theory formation and rule generation(RULEMOD). If automatic structure elucida-
in other domains where structures are found. tion is possible in the domain of chemical phe-
The rise of the transformational-generative nomena, why should it not be possible in the
approach in linguistics and the appearance in domain of social phenomena?
mathematical linguistics of the algebraic, as dis- These approaches to the conception, under-
tinct from the statistical, approach has resulted standing, analysis, representation, description
in the computerised study of the relations be- and manipulation of structures in various do-
tween deep and surface structures and the auto- mains suggest that the creation of an artificial
matic search for generative rules. In i958, Levi- intelligence system for rule formation in the
Strauss foresaw the computerised study of syn- domain of social phenomena is now a practical
tactic structures. He did not, however, consider proposition. Even when we take into account
the possibility of the computerised study of the formidable epistemological problems and
social structures. Fortes pointed out that these methodological pitfalls which will remain, and
also have a syntactic character. also the continuing personal and social sources
Computer applications and the field use of of bias, it remains a practical proposition. Col-
microcomputers are found in both archaeology laboration between knowledge engineers and
and sociocultural anthropology. In addition to experts will be necessary, but this has proved
SPSS and SP-Micro are software packages for successful in other fields (Feigenbaum I979).
archaeological quantitative analysis, such as Hugo F. Reading
AQUA and ARCHON. There are archaeolo- London
gical programmes, such as CLUSTRAN, for
cluster analysis and also programmes for pattern
recognition (Laflin I980). SOLCEM is a prog- Bertalanffy, L. von I969. General system theory.
ramme which generates an integrated overall New York: George Braziller.
interpretation of a cemetery excavation record. Blauberg, I. V. et al. I977. Systems theory: philo-
But according to Doran (I977), its limited sophical and methodological problems. Mos-
efficiency could only be improved by the ap- cow: Progress Publishers.
plication of AI work on knowledge representa- Buchanan, B. G. I979. Issues of representation
tion. For ethnographers, there are programmes in conveying the scope and limitations of
in APL for the qualitative analysis of ethnosci- intelligent assistant programs. Machine In-
ence texts and also the Electric Window prog- telligence 9.
ramme for editing ethnographic files (Werner Doran, J. I977. Knowledge representation for
I982). archaeological inference. Machine Intelli-
Research in AI has produced a number of gence 8.
intelligent systems. There are domain-specific, Ellen, R. F. I977. Polythetic classification. Man
knowledge-based inductive inference systems, (N. S.) I2: I77.
some of them pattern-directed, which seek out Feigenbaum, E. A. I979. The art of artificial
patterns in the input data. There are theorem- intelligence-themes and case studies of
provers, intelligent assistants and discovery sys- knowledge engineering. In Auerbach Annual
tems. No archaeological or ethnographic prog- 1979 Best Computer Papers (ed.) I. L.
ramme, with the exception of SOLCEM, is Auerbach. New York: Elsevier/North-
anywhere near to being an intelligent assistant, Holland.
let alone a discovery programme. Foley, J. M. I 978. A computer analysis of met-
In literary analysis, however, we have a rical patterns in Beowulf Hum 12, 7I-80.
promising development in the generation of Grenander, U. I976. Pattern synthesis. New
'metrical templates' by the automatic compari- York: Springer-Verlag.
son of successive metrical patterns in poetry Laflin, S. (ed.). I980. Computer applications in
(Foley I 978). An example of a discovery system archaeology 1980. Birmingham: Computer
limited to re-discovery is the mathematical AM Centre, University of Birmingham.
programme, which has re-discovered many Lehnert, W. I979. Representing physical ob-
concepts in number theory (Lenat I978). One jects in memory. In Philosophical perspec-

This content downloaded from 103.231.241.102 on Thu, 27 Jun 2019 08:40:42 UTC
All use subject to https://about.jstor.org/terms
 CORRESPONDENCE

tives in artificial intelligence (ed.) of connexion

M. Ringle: to ancestors; and a (residually
Harvester Press. defined) nonagnatic descent construct is which
Lenat, D. B. I978. On automated scientific at least one filial link in a chain of descent in
theory formation: a case study using the female. These three descent constructs in turn
AM program. Machine Intelligence 9. define, with reference to any salient ancestor,
Levi-Strauss, C. I972. Structural anthropology. three categories: agnatic descendants (A);
Harmondsworth: Penguin. cognatic descendants (B) (subsuming A), and
Needham, R. I975. Polythetic classification: nonagnatic descendants, (C or B-A) (i.e., a
convergence and consequences. Man category of all B's who are not A's).
(N.S.) 10, 349-69. My argument does not presuppose that all
Piaget, J. I976. The psychology of intelligence. members of category B are attributed undiffe-
Littlefield: Adams & Co. rentiated status by virtue of their being cognatic
Popper, K. R. I960. The poverty of historicism. descendants, apparently the basis of Calhoun's
London: Routledge & Kegan Paul. misunderstanding. Let me illustrate with the
Raven, P. H. et al. I97I. The origins of tax- Kwaio. Kwaio say that all cognatic descendants
onomy. Science 174, I2I0-I3, of an ancestor orioritana (lit. 'return from') that
Solheim, W. G. I976. Polythetic classification. ancestor. They comprise the category of people
Man (N.S.) II: 282. who, for instance, have rights to property cre-
Temkin, 0. i965. The history of classification ated by that ancestor. ' However, these rights are
in the medical sciences. In Classification notinequal and undifferentiated; and one basis for
psychiatry andpsychopathology (eds) M. Katz differentiation is whether descendants are
et al. agnates (futa ani wane, lit. 'born of men') or
Werner, 0. I982. Microcomputers in Cultural nonagnates (futa anigeni, lit. 'born of women');
Anthropology: APL Programs for Qual- they have different status in many contexts of
itative Analysis. BYTE 7, 7: 250-280. ritual, land rights and descent group politics.
von Wright, G. H. I97I. Explanation and under- I infer that Tallensi make similar distinctions.
standing. London: Routledge & Kegan I do not doubt that Tale individuals related to
Paul. ancestors through ba doyam (which Fortes glos-
ses as 'paternal parentage') have a qualitatively
different status, with both specialjural and mor-
Ancestors, sociology and comparative al implications, from those related to these
analysis ancestors by ma dcsam ('maternal parentage';
Perhaps Tale ancestors can at this stage best be Fortes I949: 30). I suggested that the connexion
left to rest (or exercise authority) in peace. But of Tale individuals to their ancestors (not simply
Calhoun's brief comments about my compari- their 'mothers' brothers') through non-agnatic
son (Keesing I970a) between Kwaio and Tallen- links, and the collective participation of
si conceptualisations of descent and kinship, nonagnatic descendants (Fortes's 'extra-clan
which he finds a 'forced and unsociological kinsfolk') in sacrifices to lineage ancestors,
analogy' (Calhoun I980: 309), call for clarifica- appeared to imply a Tallensi conceptualisation
tion. I do not doubt, nor did I then question, of these relationships in terms of descent, not
that there are striking differences between simply of filiation (i.e., cognatic kinship).
Kwaio and Tallensi in the structure of descent Fortes (I945: iso) himself slips into such a usage
groups. Nor did I suggest that whether conne- when he observes that a Tallensi sacrifice is a
xions are matrilateral or patrilineal [in the Tal- 'sacrament in which all the descendants of
lensi system] 'is . . . a matter of indifference' lineage ancestors, through both men and
(Calhoun I980: 3 I0). Indeed it is only a 'matter women, participate'.
of indifference' in a limited range of contexts in If Tallensi conceptualise a category of
Kwaio society. My argument had to do with the nonagnatic descendants (a category C), then
conceptual system of which descent groups are a they would seem to have to conceptualise as
realisation: that is, with the cultural categories in well a total category of cognatic descendants (a
terms of which Kwaio and Tallensi define social category B). The logic is obvious, since categ-
relationships. Since Calhoun persistently misin- ory C is residually and negatively defined. In
terprets my argument, and understands me to order to conceptualise people as descending
be arguing against 'the importance of the dif- non-agnatically, you need an inclusive concept
ferentiation between agnatic and nonagnatic of 'descending'-as implied by Fortes's 'all the
kin', a reiteration of my point seems in order. descendants of lineage ancestors, through both
Following Scheffler (I966), I noted three men and women'. Tallensi would seem to (have
types of descent construct that appeared to be to) invoke precisely such an inclusive concep-
culturally recognised among both Kwaio and tion of cognatic descent to define the total con-
Tallensi: an agnatic descent construct; a cognatic gregation entitled to take part in a sacrifice to
descent construct, in which male and female lineage ancestors. Since Tale ancestors clearly
filial links equally count in establishing a chain take an interest in nonagnatic as well as agnatic

This content downloaded from 103.231.241.102 on Thu, 27 Jun 2019 08:40:42 UTC
All use subject to https://about.jstor.org/terms