From Real Patterns to Prospective Quacks:
Review of Brainchildren: Essays on Designing Minds by Daniel Dennett
Joao Teixeira
Department of Philosophy,
University of S. Carlos,
13.565-905, Brazil
and
Center for Cognitive Studies,
Tufts University,
Medford, 02155
USA
jteixe01@emerald.tufts.edu
djte@power.ufscar.br
Copyright (c) Joao Teixeira 1998.
PSYCHE, 4(15), November, 1998.
Previously Held: http://psyche.cs.monash.edu.au/v4/psyche-4-15-teixeira.html
KEYWORDS: Daniel Dennett, Artificial Intelligence, reverse engineering, real patterns, animal mind, expert systems, medical practice, ethical issues.
REVIEW OF: Daniel Dennett (1998) Brainchildren: Essays on Designing Minds. Cambridge, MA: MIT Press, ix + 418pp. ISBN 0-262-54090-8
Price: $US20.00 pbk.
In his new collection Dan Dennett looks over a broad range of themes,
from philosophy of mind to artificial life and from ethology to animal
psychology. All the essays were written in the past twelve years and
have already been published in conference volumes or in specialized
journals. The purpose of the collection is, as the author says, to group
such essays "for the convenience of students and other readers who do
not have ready access to a major university library." The book is
divided into four parts. The first is devoted to philosophy of mind, the
second to artificial intelligence and artificial life and the third to
issues related to ethology and animal mind. There is also a final part,
which has received the title "Standing Back". This final part consists
of two essays, one providing an overview of Dennett's own work over the
last decades and another which points to the discussion of moral issues
raised by some specific uses of expert systems in medical and
professional life. This last essay was published in 1986 and it is, in
my view, the most absorbing one.
The section devoted to philosophy of mind includes ten essays, some of
which have become classic pieces, such as, for instance, "Can Machines
Think?" (with two postscripts), "The Practical Requirements for Making a
Conscious Robot" and "The Unimagined Preposterousness of Zombies:
Commentary on Moody, Flanagan and Polger". The first is a lengthy
commentary of the significance and prospects of the Cog project
developed by Rodney Brooks and his team at the MIT. Dennett presents
several philosophical implications of designing a humanoid robot as the
realization of the major thought experiment of this century. The last
essay of this section, "The Unimagined Preposterousness of Zombies:
Commentary on Moody, Flanagan and Polger", appeared in the Journal of
Consciousness Studies in 1995. It contains an attack on the ontological
possibility of creatures such as zombies. Zombies would be behaviorally
indistinguishable from human beings and, still, lack "qualia" in so far
as their inner lives would be "informationally sensitive" but not
"experientially sensitive" like ours. The attack is purported to be an
utmost rejection of a current realism concerning "qualia" as well as
sharp criticism of the claim that an account of the nature of
consciousness involves extra ingredients beyond functional organization.
It is interesting to read these two essays one after the other, since
they suggest an implicit order of ideas: disavowing extra ingredients
for consciousness besides functional organization and "qualia" seem to
be the necessary steps to conceive of the possibility of designing a
conscious robot. However, the issue concerning "qualia" is still being
debated. Much has been written about the nature of "qualia" and their
relationship to consciousness, zombies and "zimboes" in the past few
years but, still, the debate concerning such epiphenomenal ghosts is far
from being settled once and for all. Such a debate reveals of a lot of
misunderstandings about what we mean when we talk of "qualia", "inverted
qualia", "absent qualia" and so on. This is the motivation of Dennett's
essay "Instead of Qualia" in which, however, he confesses that he has
lost the battle to reach at least a clear and agreed-upon meaning for
the word. So viewed, the philosophical debate is likely to be endless
and, most likely, not to reach even the minimal semantic consent
required to answer the ontological questions surrounding the issue posed
by the existence of "qualia".
No doubt the most important and controversial essay reprinted in the
section devoted to philosophy of mind is "Real Patterns" where a whole
reassessment of Dennett's brand of functionalism is at stake. In this
paper of 1991 Dennett attempts the difficult task of steering between
the Scylla of instrumentalism and the Charybdis of realism. The essay
has already received critical commentaries from John Haugeland (1993)
and, more recently, from William Wilkerson (1997) but several issues can
still be raised about Dennett's variety of realism concerning mental
states and, more particularly, folk psychology.
Haugeland bases his overall criticism of Dennett's notion of pattern in
the need to disambiguate between two kinds of issues. On one hand there
is the sheer ontological commitment to the existence of the elements
that compound a pattern. On the other hand, there is a further and more
forceful commitment to the existence of the pattern itself, that is, to
something beyond such constitutive elements. The first can be
trivially ascertained, the second involves a certain brand of
ontological holism. A vacillation in distinguishing those two levels of
analysis and those two notions of pattern is, according to Haugeland,
pervasive in Dennett's account and also responsible for several
difficulties concerning the kind of reality we are to ascribe to
intentional entities. But I shall not pursue Haugeland's criticism here.
Rather, I shall briefly discuss Wilkerson's claim that an account of the
reality of the patterns would ultimately have to be sought by
establishing some mapping between folk psychology and brain structure.
Beliefs and desires are a paradigm case for such a discussion since they
are the most common elements of folk psychology involved in the
intentional stance. In accordance with Dennett's intentional stance,
beliefs and desires describe and predict patterns of behavior, that is,
they describe an objective reality. In other words, such patterns are,
according to Dennett, real, if two criteria are taken into
consideration. First, a pattern is real if a "compression algorithm" can
describe such a pattern, that is, if there is a description taking
advantage of some basic regularity. Second, a real pattern occurs if
there is predictability, that is, if predictions are successful.
Wilkerson's contention runs as follows: on the one hand folk psychology
can be considered instrumentalistically, since beliefs and desires are
posits or abstracta in the same way as centers of gravity or other
useful constructs involved in the explanation of a specific natural
phenomenon--in this case, human behavior. On the other hand, it may be
the case that beliefs are something in the head, since for Dennett minds
are brains. So one could ask whether or not such theoretical entities
would ultimately map to brain structures and, in this case, a tension
between instrumentalism and eliminativism (and realism) would be
inevitable. But there is more to Wilkerson's claim than questioning the
tension between instrumentalism and eliminativism. Wilkerson seems to be
pressing a more forceful concern, namely, the need to find "how the
various states and entities of folk psychology are realized in the human
brain" (p. 558). In order to make sense of the reality of the patterns,
Dennett's approach would have to be supplemented with an account meant
to bridge the "gap between a mechanistic, non-intentional description of
the brain's activity and the intentional description of behavior that
the brain must somehow make possible" (p. 562). In other words,
Wilkerson attempts to provide an account of how patterns could possibly
be generated by the brain, that is, how "the pattern in human behavior
is reflective of a real pattern realized in the structural features of
the brain" (p. 562). He rejects the hypothesis that patterns could be
generated "by a program that manipulates various symbols (tokens of a
language of thought) by virtue of the symbols' syntactic features" (p.
562). In so far as Dennett would also consider such a hypothesis
implausible and way ahead of empirical data, he proposes a pattern
generator based on parallel processing and randomness, namely, Dennett's
Joycean Machine. By transforming multiple channel information into
serial processing the Joycean Machine would account for the production
of patterns and would also provide the possibility of drawing a
connection between two levels of description: the description of the
brain's activity (Dennett's physical stance) and the intentional
description (Dennett's intentional stance). Besides overcoming the
tension between instrumentalism and realism, Wilkerson suggests a
further gain in his reconstruction of Dennett's notion of patterns:
avoiding the possibility that Dennett's view might ultimately clash with
the hopes of contemporary neuroscience, that is, the hopes for a
possible reduction of folk psychology to some kind of brain
structure.
I do not believe that Wilkerson's contention of an expanded account of
Dennett's view of the reality of the patterns can help resolve any
tension between instrumentalism and realism--if there is such a tension
after all. Dennett's approach leaves open the possibility that such
abstracta may or may not map onto something in the brain. I do not
suppose that Dennett would reject a possible mapping between folk
psychology and some brain structure, despite his claim that such a
mapping would be subject to indeterminacy. Furthermore, the adoption of
the intentional stance does not mean that any attempt to unravel the
neural correlates of mental phenomena should be abandoned. The issue
does not bear on the adoption of either instrumentalism or realism (or
eliminativism) as a resolution for an alleged tension in Dennett's
approach--such labels fall short any proper characterization of his
position. For Dennett's presumption in defining his position--a position
he would like to call mild, intermediate or semi-realism--is nothing
over and above a commitment to naturalism. Naturalism does not require
eliminativism--not in the sense that patterns are to be reduced to brain
structure in order to be real. So viewed, the ontology of such posits
or abstracta does not seem to be one of Dennett's major theoretical
concerns. Folk psychology and its patterns will continue to be as real
as centers of gravity are whether or not their possible reduction to
brain structure can be accomplished. The reality of patterns is
supported by their usefulness in as much as they provide a shortcut for
our cognitive apparatus to deal with the unpredictability of the
behavior of some agents in the environment. Folk psychology may continue
to prevail in our day-to-day explanations and predictions of behavior
for its pragmatic/heuristic advantages provided by compression
algorithms. It is very unlikely that folk psychology will, some day, be
replaced or supplemented by more sophisticated concepts derived from
neuroscience. As Dennett remarks, "A truly general-purpose, robust
system of pattern-description more valuable than the intentional stance
is not an impossibility, but anyone who wants to bet on it might care to
talk to me about the odds they'll take." (p. 120). So viewed, the
alleged clash between Dennett's instrumentalism and his realism (or
eliminativism) seems to arise from a misinterpretation of the purposes
of his theory. For how patterns could be generated by the brain or by
some kind of virtual machine is a subsidiary issue--one that cannot
substantially alter Dennett's account.
Let us now turn to the second part of Dennett's book. This part groups
together essays on Artificial Intelligence and Artificial Life. At least
two of them must be mentioned: "Cognitive Science as Reverse
Engineering: Several Meanings of Top-Down and Bottom-Up" and "When
Philosophers Encounter Artificial Intelligence". The first draws a
pathway between Artificial Intelligence and biology and evolution
through combining top-down and bottom-up strategies--a task where
Artificial Life plays a prominent role. The second essay focuses on the
relationship between Artificial Intelligence and Philosophy, that is,
how those two disciplines can possibly be intertwined. This is not a
vexed question, nor a platitude as many may suppose. Many philosophers
still tend to confine themselves and look at science and at Artificial
Intelligence with some form of suspicious contempt. A fearful disdain
arises in so far as many questions addressed by Philosophy are also
addressed by Artificial Intelligence (What is mind? What is meaning?
What is rationality?). Dennett's approach to this question is to show
that it ultimately boils down to the issue of whether of not we are to
take mind as an object of scientific study. However, such an approach
cannot be fully attained unless philosophers relinquish some of their
current prejudices concerning the dignity of the ultimate mystery, the
Inexplicable Mind--a kind of pride which still lingers among
philosophers. The issue is summed up in a nicety: the prejudice against
the study of mind as an object of science is comparable to Aristotle's
prejudice against extending earthly physics to the heavens.
Besides the two essays mentioned above, section II provides at least two
more papers of historical interest, both written in a fairly agreeable
style. "The Logical Geography of Computational Approaches: a View from
the East Pole" avails to the reader a wealth of information about the
development of Cognitive Science in the eighties. Of no less importance
is Dennett's detailed review of Unified Theories of Cognition, where a
thorough appraisal of the SOAR architecture and Newell's research
program for Cognitive Science is presented and discussed.
The section devoted to Ethology and Animal Mind contains one of the most
provocative essays written by Dennett in this field: "Animal
Consciousness: what matters and why". Most significant is the fact that
this essay was published in the fall of 1995 when his book Kinds of
Minds was probably in its penultimate draft. The paper advances several
views that were to appear in the book; views that still may stir
indignation and suspicion among animal lovers. But surely this is not
the purpose of the essay, unless one sticks to some biased
misinterpretation.
No doubt Dennett's view on animal consciousness disavows some orthodox
conceptions such as those supported by Thomas Nagel (1974) in his
classic paper "What is it like to be a bat?", and, more recently, by
Mary Midgley (1984) and Peter Singer (1981). However, the purpose is not
to disqualify animal mind, nor to downgrade their suffering so as to
generate a comfortable setting for hunters, farmers and experimenters.
Rather, the purpose is to dismantle the myth that we can discuss animal
consciousness (if there is any) by projecting on such creatures our own
current conceptions of mind, consciousness and pain. The problem with
such views lies in the fact that they are deeply rooted on the
presupposition of the existence of a Cartesian Theater where conscious
processes occur, or, in other words, on the assumption of the further
substrate of a res cogitans besides brain-and-body activity.
Furthermore, as is emphasized in Kinds of Minds, one must distinguish
between mental competences and mental lives, that is, resist the
habitual conception that any mental activity is accompanied by a stream
of consciousness. It is this natural inclination, followed by authors
such as Nagel, that renders the issues involving consciousness and also
animal consciousness an impenetrable mystery or an insurmountable
philosophical conundrum from which several ethical dilemmas may emerge.
By dismantling the mythical presumption of the existence of such a
philosophical ego and its projection onto either ourselves or another
species, one is likely to get a better footing to overcome such
conceptual difficulties. This task, however, may require several
previous steps, such as, for instance, reversing some acquired
conceptions according to which animals feel and perceive the world in
accordance to patterns that underlie our own organization of the world's
experience. Such informational unification "is not anything we are born
with, not part of our innate "hard-wiring" but in surprisingly large
measure an artifact of our immersion in human culture" (p. 346).
Furthermore, in support of this idea, Dennett says that "My claim is not
that other species lack our kind of self-consciousness as Nagel and
others have supposed. I am claiming that what must be added to mere
responsivity, mere discrimination, to count as consciousness at all is
an organization which is not ubiquitous among sentient organisms" (p.
347).
This is a fascinating and original view of the nature of animal
mind--one that emerges from disentwining an item of popular
philosophical mythology deeply ingrained in our current intuitions about
the mental life of other species. Examples in support of such a view are
elegantly chosen and presented by Dennett throughout the essay. The only
noticeable theoretical difficulty with such a view is that it leads to
the postulation of unconscious pains--an undesirable consequence
which is, nonetheless, acknowledged by Dennett in his attempts to
conceive of animal pain as somewhat different from human pain. Apart
from that, our current intuitions about animal minds can still be fairly
tricky. Those who were brought up in rural areas are likely to be
familiar with the gloomy scene of a pig when it is put to death. Any
such recollection will invariably include the animal's desperate bawling
when it is captured in the pigpen--as if it knew or was
aware beforehand of its hopeless fate. This is not saying that
Dennett's view is to be rejected but rather pointing to the huge
difficulty posed by any attempt to reverse our current cultural
constructs regarding the notion of consciousness. How far can we
undertake such a task without incurring some unacceptably
counter-intuitive claims?
The last part of Dennett's book contains only two essays, but at least
one of them requires a more lengthy commentary since it addresses some
paramount ethical issues emerging from the relationship between human
beings and computers. This piece, entitled "Information, Technology and
the Virtues of Ignorance" originally appeared in Daedalus in 1986. The
essay points to the dangers of the progressive replacement of human
beings by computers in tasks that involve ethical responsibility, such
as, for instance, medical diagnosis and prescription of medical
treatment. Although expert systems for medical diagnosis had only an
ephemeral pinnacle in the seventies, the substitution of "artful
investigation" by a technology of medical diagnosis is still on the way.
The traditional medical practice as relying mostly on the physician's
ingenuity and capability will, sooner or later, incorporate the use of
expert systems and, at the same time, pose the issue of the delegation
and transfer of a great deal of human responsibility to some specific
technology.
In the course of this essay, Dennett seems to alternate between
fascination and concern as to the use of expert systems in medical
practice. Fascination comes in so far as the author does not suppose
that the partial replacement of the personal (and personable) physician
can constitute a good reason to forestall the advance of technology--we
cannot bet on the "virtues of ignorance". Such a replacement can, at
worst, be viewed as a change of social roles, as happened in the case of
calligraphers, potters and tailors who enjoyed a much more eminent
status in their communities many years ago. Nostalgic feelings that may
arise in the future are too flimsy a reason to prevent the use of new
technologies. However, concern crops up in so far as in the use of new
medical technology there is the implicit compliance to ceding
responsibilities: to what extent can we rely on expert systems for
medical diagnosis and treatment? Who are we going to blame in the case
of fatal errors? There is more to this issue: the replacement of the
traditional physician--illustrated by the figure of the rural
doctor--involves the substitution of the traditional medical interview
by the use of computer designed questionnaires. Doctors will tend not to
deal personally with their patients. A lot of excitement and
respectability, which is normally associated to the physician's social
role will diminish. Furthermore, a career opportunity may lose its
traditional enticement. In this case, says Dennett, "information
technology, which has been a great boon in the past, is today poised to
ruin our lives--unless we are able to think up some fairly radical
departures from the traditions that have so far sustained us" (p.
368).
The issue posed by the advent of expert systems in our societies does
not affect only the physicians, but all careers and all aspects of human
life. We are seeing not only of the replacement of manual work, but also
of mental work. Besides, the common sense expectation that technology
would increase our leisure time had also to be abandoned. Surely this is
not what is going on in our societies. Still, Dennett is right in
suggesting that the solution for this issue does not lie in a radical
forestallment in the advance of technology. No one could reasonably
agree to such a view nowadays. The problem does not seem to be how to
conceive of technology, but rather how we can reasonably conceive of
ourselves before the technology we produce--a technology that forces on
us the need of a thorough and quick revision of many philosophical
conceptions, including ethical ones. Philosophy seems to have
disappointed us as far as those issues are concerned. This is echoed in
Dennett's essay when he points to the need for "moral first aid" and
asks "What to do until the Doctor of Philosophy arrives"? But will the
arrival of the Doctor of Philosophy make all that much difference? How
quickly can she or he point to solutions? In other words, can human
reason provide some solution to the ethical problems stemming from the
production of human technologies?
An alternative to the Doctor of Philosophy is to expect that technology
can contribute to pointing out solutions to problems generated by its
own use. That would also include possible new emerging ethical issues.
But such a hope seems to entail a confrontation of human reason with
itself. Curiously enough, contemporary technology seems to have been
prolific in devising experiments to test the proclivities and power of
human reason. This is the case, for instance, of the celebrated
chess-playing computer program Deep Blue. Most of the time, when we
speak of a confrontation between human intelligence and machine
intelligence we forget we are the ones who designed machine
intelligence. We want to make sure that human intelligence can
ultimately supersede itself in the form of a technology that,
nevertheless, we have invented. Now, to what extent can human
technology offer alternatives to the problems posed by its own
development? Can an answer to the ethical problems posed by the medical
expert systems be supplied by Artificial Intelligence? Can we overcome
ethical and moral dilemmas by designing an expert system to deal with
such matters--a "moral first aid kit"--while we wait for the Doctor of
Philosophy to arrive?
This is not specifically the question Dennett addresses in his essay,
but I do not suppose he would be totally averse to discussing this
latter alternative. But he could point to several obstacles to the
realization of such a project. The most difficult to overcome is that it
is not possible to foresee all the consequences of human action, let
alone its ethical consequences. Dennett's view of ethics as a possible
scientific discipline--at least as far as this matter goes--would model
it as sharing all the uncertainty of meteorology. Like weather
forecasting, the possibility of predicting the consequences of human
action and its possible ethical consequences may be an almost
insurmountable problem for the design of an expert system devoted to
moral decision-making.
There is also a further question not addressed by Dennett in his essay.
On one hand, expert systems may improve medical practice by avoiding
possible malpractice suits. However, on the other hand such a technology
may lead to a surreptitious downgrading of the standards that would
normally be demanded in a medical school. This is not to say that
medical schools will collapse because of expert systems. However, as a
natural trend, we tend to spend less time studying what we know a
machine can do for us--sometimes much better than we do. This is a
possible perverse side effect of the use of such a technology--a side
effect whose consequences can affect medical education. If this is the
case, doctors will be less prepared. Not all doctors, though. There
will be an elite of expert systems designers who will keep mastering
almost all of medical knowledge. This is an even more dangerous risk
stemming from the use of this technology, that is, the risk of widening
up even more the present educational differences. The melancholic
pre-technological "virtues of ignorance" would turn out to be a real
danger of ignorance.
Will the use of expert systems in medicine imply that our future doctors
(except for the elite we referred to above) will be quacks? This is
likely to be true, at least as far as a first generation of doctors is
concerned. However, we do not know exactly how such a relation between
doctors and expert systems may develop in the future. It might be the
case that such an elite itself will also disappear if medical knowledge
expands to a point where no human doctors can have the mastery of all
medical information. New kinds of expert systems will have to be
designed; in this case expert systems to design other expert systems
and, in so doing, summarize and keep up-to-date knowledge that would
normally be attained by human specialists. If this is the prospect, even
the elite will disappear and a new way of conceiving of knowledge and of
knowledgeable people will have to replace our ordinary views.
There is much more I could say about Dennett's book, a collection which
expresses twelve years of his intellectual trajectory and whose
variegated approach to so many subject matters makes for fascinating and
entertaining reading.
Acknowledgments
The author is supported by FAPESP grant # 97-03518-6.
References
Dennett, D. (1996). Kinds of minds. New York: Basic Books.
Haugeland, J. (1993). Pattern and being. In B. Dahlbom (Ed.) Dennett and his
critics (pp. 53-69). Cambridge: Blackwell.
Midgley, M. (1984). Animals and why they matter. Athens: University of
Georgia Press.
Nagel, T. (1974). What is it like to be a bat? Philosophical
Review, 83, 435-450.
Newell, A. (1990). Unified theories of cognition. Cambridge, MA:
Harvard University Press.
Singer, P. (1981). The expanding circle: Ethics and
sociobiology. Oxford: Clarendon.
Wilkerson, W. (1997). Real patterns and real problems: Making Dennett
respectable on patterns and beliefs. The Southern Journal of
Philosophy, 35, 557-570.
