Summer 2011
Thursday, 07/14
Machine Functionalism
• Claims that the mind is a (very complex)
computer program.
• One that arises naturally, not one that’s
designed or programmed by anyone!
• But what exactly does that mean? How can
we understand this in a concrete, nonmetaphorical way?
Machine Functionalism
• This is where machine tables come in.
• A machine table is a complete set of
instructions for the machine’s operations, or a
specification of the machine’s software.
• You can think of it as a complex “job
description” for a machine. The job
description for a heart is to pump blood, but
the job description for a coke machine is
captured by the following table:
Go into
this State
this Output
Machine Functionalism
• The machine table for the coke machine captures what a
coke machine essentially is. Any device that implements (or
works according to) the machine table counts as a coke
• Similarly, the (far more complex) machine table for the
mind captures what a mind essentially is. Any device (e.g. a
human brain, a Martian brain, commander Data) that
implements such a machine table counts as being (or
having) a mind.
• Individual Mental States (e.g. pains, hopes, desires) are just
states of the incredibly complicated machine table for the
mind. To be a pain is to be a state that responds to inputs,
delivers outputs and interacts with states in a way that
accords with one of the machine table states.
Machine Functionalism: Problem
• Machine states are parts of the specification of the job
description of an entire system, e.g. a coke machine or a
• Different machines with different machine tables (or
total “job-descriptions”) cannot have any states in
• Since human minds would be captured by different
machine tables or total “job descriptions” than animals,
humans cannot share any mental states with animals (or
other creatures).
• So machine functionalism seems to drop one of the main
motivations for going functionalist in the first place.
Causal-Theoretical Functionalism
• A more general version of Functionalism that
doesn’t rely on (Turing)machine tables.
• What makes a pain a pain (or, more generally,
any specific mental state the mental state it is)
is its having a certain causal role, i.e. its being
caused by some sensory inputs or mental
states and causing behavioral outputs or
mental states.
Ramsey-Lewis Method
• A way of characterizing individual mental
states (e.g. what it is to be in pain, to
want ice-cream, or to believe that it will
rain tomorrow) in a non-circular or
regressive way, without using
(Turing)machine tables.
• Unless the functionalist does this, her
theory fails to be a reductive theory of
mental states, or characterize mental
states in non-mentalistic terms.
Ramsey-Lewis Method
Here’s basically how it works.
Step 1. Write down the job-description of pain
in the form of a theory:
Pain Theory: ...and pain is caused by pin
pricks, and pain causes distress and wincing
and avoidance behavior, and distress in turn
causes brow-wrinkling...
Ramsey-Lewis Method
Step 2. Form a “Ramsey sentence” (by
replacing all the occurrences of mental states
by variables).
Ramsey sentence:
There are states S1 and S2 such that S1 is
caused by pin pricks, and S1 causes S2 and
wincing and avoidance behavior, and S2 in
turn causes brow-wrinkling…
Ramsey-Lewis Method
Step 3. Use the “Ramsey sentence” to characterize what it is
to be in pain (or to be in distress) without presupposing any
prior understanding of pain or other mental states:
A person is in pain =def There are states S1 and S2 such that
(S1 is caused by pin pricks, and S1 causes S2 and wincing and
avoidance behavior, and S2 in turn causes brow-wrinkling…)
AND the person has S1.
A person is distressed =def There are states S1 and S2 such
that (S1 is caused by pin pricks, and S1 causes S2 and wincing
and avoidance behavior, and S2 in turn causes browwrinkling…) AND the person has S2.
Varieties of Functionalism
• Analytic functionalism.
The causal job-decriptions of mental states
derive from our concepts.
• Psychofunctionalism.
The causal job-decriptions of mental states
are to be discovered by empirical
A Priori vs. A Posteriori
Analytic Functionalists claim that the
specifications of the job-descriptions for
mental states are knowable a priori.
Psychofunctionalists claim that they are
knowable a posteriori.
A priori = with justification independent of
A posteriori = with justification that derives
(in part) from experience.
Role vs. Realizer Functionalism
• Realizer state for pain: the internal state which realizes
the pain-role (or job-description), in a given creature.
• Role state for pain: the “second-order” state of having
some internal state or other that realizes the pain-role.
The role functionalist identifies pains with role states.
The realizer functionalist identifies pains with realizer
states. So realizer functionalism is really just a
sophisticated version of the identity theory.
Problem for Role Functionalism: The Causal Efficacy of
Abstract, High-level properties.
Necessary vs. Sufficient Conditions
• X is sufficient for Y if when X obtains, Y always
follows (i.e. when the presence of X
guarantees the presence of Y).
• X is necessary for Y if, should X fail to obtain, Y
cannot be the case (i.e. X must be satisfied for
Y to obtain).
The Turing Test
• Original version: Interrogator (in one room), a
man and a woman (in another room),
communicating via a teleprinter. The interrogator
must determine which one is the man and which
one is the woman. The woman is sincere but the
man is a deceiver. Test: could a computer play the
role of the man as successfully as the man?
• Equivalent test. Can a computer fool a human
interrogator (communicating with it via typing)
that it is a man?
The Turing Test
Discussion Q’s:
Q1: Is passing the Turing Test sufficient for
genuine thought and intelligence?
Q2: Is it necessary?
• What is Machine Functionalism and how is it
different from Causal-Theoretical
• What’s the difference between Analytic
functionalism and Psychofunctionalism?