COMMENT ⊗   VALID 00009 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	motiva[f81,jmc]		On human motivation, 1981 dec
C00008 00003	moved from MOTIV.MEN[F75,JMC] 19-Apr-76	Brief note on motivation
C00009 00004	from MOTIVA.MEN[F75,JMC] 08-Dec-75 ON MOTIVATIONAL STRUCTURES - ESPECIALLY HUMAN
C00014 00005	MOTIV2.MEN[S76,JMC] 02-May-76	Peculiarities of human motivation
C00016 00006	del motiva.2[f81,jmc]
C00019 00007
C00025 00008	How misunderstandings about human motivation lead to harm
C00029 00009	Redesigning our motivations
C00033 ENDMK
C⊗;
motiva[f81,jmc]		On human motivation, 1981 dec

	Most of the trouble in society comes from the fact that
human motivational structure is not well enough suited to human society.
There are several problems.

	The first problem is that our long term motivations and our short
term motivations often don't agree.  A person may have a long term goal,
and by long term we can mean anything from an hour to beyond a lifetime,
but his actions on behalf of the goal depend on his being immediately
motivated to do what is required.  The old-fashioned folk psychology
characterization of this problem is that people often lack will-power.

	This problem is to be expected considering our evolution from
ape-like ancestors or even from primitive man.  Our ancestors had no
opportunity to form very long term goals.  Apes don't seem to have
them.  In fact perhaps no animal other than a human forms and executes
long term plans.

	In order to explain this we need to consider why squirrels'
storing nuts for the winter doesn't count as a long term plan.  A
squirrel doesn't know about winter.  Presumably a young squirrel
born in the late Spring and separated from other squirrels immediately
after birth will still store nuts starting in the Fall.  Of course,
this can be regarded as executing some kind of instinctive plan, but
the human capacity we are considering is different.  It is the ability
to think about a desired future, make a plan to bring it about, and
then perform actions whose only motivation is that they further
the plan even if the actions are unpleasant to perform.

	There is no reason to expect that a creature recently evolved
from the ape would be very good at this, i.e. would be good at
actually carrying out plans at the cost of short term unpleasantness.

****

	A second consideration is that we have goals about our own motivations,
and these goals are often difficult or impossible to realize.
In fact they are usually so imprecisely formulated that it is often
impossible to tell whether they have beenreealized or even to concretely
envisage their realization.
We may make an analogy with metamathematics and call them meta-motivations,
but little that I shall say requires knowledge of this analogy, although
it played a role in leading to these ideas.

	For example, we would like to be rational.  An extremely rational
person would have an ordering on possible world histories or bettter
yet an evaluation function, and would compute and follow a strategy with
the highest expected evaluation.  A simple form of evaluation for a person
is a sum or integral of the events that "happen to this person" which
he tries to maximize over the remainder of his life.

	A behaviorist might simplify further by considering the sum to be
over sensory inputs.  However, this wouldn't account for people attempting
to achieve goals that can't be realized in their lifetimes - except in so
far as these could be re-interpreted as the goal of wanting the experience
of expecting the grand plan to be realized.

	One strategy for optimization is to optimize for each day or
month though this leads to a general optimization only under very
special assumptions.

	As a second example, a person might like to be good, i.e. to
be motivated by the welfare of others.

	A program could be designed so as to want to change its
motivations.  According to the design, this might be easier or harder than
for humans.
moved from MOTIV.MEN[F75,JMC] 19-Apr-76	Brief note on motivation
In my opinion, the main intellectual qualities of the human mind
are determined by the nature of the problems we face.  Any other
being that can solve problems would have similar intellectual
qualities, and we will have to program similar qualities in order
to make machines behave intelligently.  To take a simple example,
no machine or animal can play chess well without some tree search.


On the other hand, the human emotional and motivational structure
is very much determined by the particular way our species evolved
- our family structure, 
from MOTIVA.MEN[F75,JMC] 08-Dec-75 ON MOTIVATIONAL STRUCTURES - ESPECIALLY HUMAN
ON MOTIVATIONAL STRUCTURES - ESPECIALLY HUMAN

	While the belief structures of machines are similar to
those of humans, because they are determined by function, the
motivational structure of humans is quite different from any that
we would be likely to put in a machine except for the explicit
purpose of simulating a human.  One might wonder why this is so,
since one might expect the motivational structure to have been
subject to selection pressure just as the intellectual structure
has been.  Well, apparently it is being subject to selection
pressure, and many variants of human motivational structure
are being deselected by failure to reproduce.

	We begin by considering machine motivational structures:

	#. The best understood is the simple servo-mechanism.  A
certain function measures the deviation of the present state of
the system from a desired state, and generalized forces are applied
to the co-ordinates of the state in such direction as will reduce
the deviation.  A common elaboration is the multi-loop system in
which the behaviors of parts of the system are controlled by
separate servo loops.  The most common malfunctions of such systems
involve instabilities caused by time lags the wrong amount of gain
or the wrong amount of damping.  The human contains a number of
such servomechanisms, but the overall motivational structure does
not have this form.  I am somewhat doubtful whether terms like
%2wanting%1 can usefully be applied to servo-mechanisms.

	#. The second mechanism is a tree search problem solver

	#. Beyond this we come to goals expressed in terms of incompletely
axiomatized abstractions.  We may try to give a machine the goal of
advancing the welfare of humanity without fully specifying what
the welfare of humanity consists of.  This is ok and even safe
provided explicit prohibitions of certain kinds of actions are
included to avoid cses wherein elaborate reasoning with insufficiently
understood concepts gives bad results.

	#. The ultimate in rationality orders probability distributions
of histories of the world, and asks the machine to optimize.  Unfortunately,
histories of the world are unknowable and the probability distributions
are uncomputable.  Moreover, we humans can't do it either, so we wouldn't
know what program to give the machine in any concrete sense.  Nevertheless,
the machine could be told to optimize the probability distribution of
future histories of the world and given such partial knowledge of the
world as we can formulate.  It must also be given some rules for
making arbitrary choices when the explicit optimizations don't give
a definite answer.  Of course, it must know when it is making an
arbitrary choice, so it can take it back if more definite information
becomes available.

MOTIV2.MEN[S76,JMC] 02-May-76	Peculiarities of human motivation

~	Human motivational structure has the following characteristics
which don't seem necessary for problem solving and which are not
shared by any present machines:
.subitem←0;

&. Strength of motivation has a quantitative character.  One can decide
to do something but not have the necessary strength of will.

&. When a subgoal is created in support of a main goal, it can
escape from the main goal and remain in force even when it no
longer contributes to it.

&. Motivations are modified by the physical condition of the body, e.g.
by tiredness.  It is not ti{w79ε{;∃βO→βS'⊗+⊃1β}s∃β←∞sSMβ⊗+OQ1ε∪WP4W;#πQπ≠↔↔7~β←?K&C←#'f)β'9εc'≠∃εMβ¬π;#?3*β'Mβ∞3≠↔∂&+⊃βJβS'K.#;↔O~βπ;⊂hS?S#/⊃βC#O≠'∂πbβOSπ&+M84Ph)→9∧kW∂!εCW7πrβ7?SO3πS'}qβ'Mε+C'O}#'
mε;'7∞aβ7?&K[πSN{9β'~β↔[↔rβ7?K(h+O=ph(4(LIβ#π6)β#↔∂∪⊃βπ⊗;W7↔w#MβSFQβSF)βπ␈3∃β≠.SWK/→β?→εCW7πph+7?&K[πSN{9β7∂Iβ∃ε+OO↔w#'π1ε3?IβFK∨!βf+[↔1εK;S↔fc↔∂S.1β≠.s∂S'}sM04V∪WQβ6+]β∂}s∂K↔&)βπK?+7↔;'→9↓α≡K;∂∃ε{→βSF+O∃β∂≠C↔∂'→β?→εCW7πrβ7?SO3πS'}p4+π⊗)β↔π≡K3eβ/CC3πNs↔⊃β∂→β↔[}cWS'}sπKeπ≠WK[O3π3Mbβ'Qβ>{W3⊃ε∪∃β≠␈∪SW'&{WL4VK→βSF+eβ←/∪∃β+/≠Qβ←FQβ←∂→β;↔.#↔⊃β6{Iβ#N;!β3/3↔1β∞≠S'[O#'↔MqP4(3∪↔bβ7?SO3¬9J↑1aE3Vk∞t4T≠?7Cf+c'SJβ?→βn{S'[∂#'?9ε;⊃βε{3'SN≠L4(hP&S#*β←?Kf!β#π~βS?=εkπ;eεOC↔∨#Mβ≠␈⊃β¬βG+7π9π#=β#∂3∃β¬π≠';∨f(4+W&K3'SJβ≠W;∨#'?9π##πQπ#π/↔~β';Szβπ∂∂␈+;Qβ∞c1βπ∨β↔∂S~β?→β&C∃β←␈∪3⊃8hRS#↔⊗+≠?K*β←∃β'KC'∂∞c3eβ∨;'S∂Bβ?WIεSS↔w#'?9ε{[↔Iπ3πKgNs≥βC/∪'?∪_h+π7}s≥β∂}c3↔∂&K?;Mε{→βπ∨β↔∂S~β?→β&C∃β←␈∪3⊃9αα?WIεSS↔w#'?9εKMβ?7#↔84W≠←'S≡C↔⊃β↔IβKπv#?5β/3↔;S~βS#π"β7π/*β?;∃ε{Iβπv{S#↔∩βπOC.≠Qβ?2βS#∀hS←?Kf!β7?n+;Sπ⊗K3eβ≡3'↔w!9↓α&C∃β↔∨≠↔;∂*β?→βε{3'SN≠Mβ'~βS=βn/∀4W≠π3'.sQβSF{O∃β∂≠C↔∂'→β?→π##∃β>{K3⊃ε?W"β←#'≡AβS#*β↔3↔∨#?Kπ&(4+∂∞qβ#?ε)βS=ε∪↔;↔6KQβJβ∨'[Ns≥βSF)βC?fKS'∂N9βC␈;↔I8hP4*'~βS#↔⊗)β;?"β¬β∨}1β?2β#π[Ns≥β∨}3M|hP4(&⊗+O'∪/→β∂?oβ3↔aε;?π1n#'K↔∨#↔⊃β⊗+#π[N{I1β>)βπ3≡yβ';'+3∨∀hS'9β⊗+≠3↔FK[∃β⊗+#π[N{I1β↔+QβO}k↔S'n+MβSF)βK↔6c↔aβ∞≠S'?rβ'Mβ&C∀4+'∪'∨∨/∪';≥ε{→β∂}kC3↔Bβ↔#∂3'?Iph(4*⊗+∂πW≡)β←∃εCπ[∃π≠W∂!πβ??Iπ≠#?K"βS↔Kjβ7↔7␈∪'↔Mbβ←∃β/≠∃βSF)β↔c&+K;π`h+←?⊗c⊃β≠␈⊃β7↔n{Keβ∂→β7W≡AβπMπβ?OON∪3∃9ααS#W~β←∃βπ∪?∨K∞iβ?W↔≠↔3[/_4+OzβS#π"βS#∃εs↔cQπ≠S↔AεK9β¬ε≠?7Cf+aβ.Cπ['␈⊃βO↔∂+↔;∂*β'Mβ'∪'∨∨/∪↔⊂4V∪eβSF)β∂?oβ3↔SN{9β?2βS#∃πβK↔[N{WMβ∨#↔A↓jβKπSF+IβSF9β.K;≥β⊗+7↔7⊗+K↔⊂hSπMβ&C∃β;/CQβO&+A9↓∧∂SW∞c3eβ&C'MβO→β#'∨#?K'≡33eεK;↔c∞≠Q1β≡K;∂∃ε;'7∞`4+.Cπ['␈⊃β'MεkW∂!εk?K∃π#K'∨>+K↔⊃ε∪eβOO#WπSN{;M9ααS#∃εK;;?6S'?rβ'Mβ&CπP4W;∃βO}k↔S'n+Mβ←␈∪-β←O#!β'n∨';.!βO''+πS'}sM84Ph*W;&+Iβ←FQβ∂}s∪'SN{;Mβ&{↔Mβ␈βS'7OS';≥ε	βO↔∂+↔;∂*β?→β∂≠C↔∂'→β?→ε	βO''+πS'}p4+K/≠W3QεK9β?6+Kπ3bβ#'3bβ∂3'n∪';≥z↓αS#O→βGW/≠S'?rβπK'≡+Mβ.≠πWO(h+C?fKS'∂~βO↔↔o→βS=εK;[?g3∃βO>KS∂#Ns≥βπ'#↔;SN{9β≠⊗{5β?v)βπOε+∂Qβ}04+SF)β∂?.sSKe?→β←↔f3πK∃π#=βπv{S#↔∩aβ↔≡WO∃π##∃βε{3'SN≠π1β∨KOS↔jβ#πLhS↔[↔rβ3↔O~β∂πC∞≠'Seπ##π9ε	βC↔↔≠?9β}1β/↔/β';≥π≠↔[↔⊗1βSFK;∨MεK9β7Ns⊂4+∂!β?;≡)84(04*n{K∃βv{S↔Mε{9β7␈#'[π&K?98hP4(&∂→β←∃εCπ[∃πβ?';&+⊃β?/!β↔6{K∃1εCW7πrβ7?SO3πS'}qβ≠3.≠SWπ&+M9↓∧λ4+#.kπ9βneβ←O≠!β≠␈⊃βπ;O##';:βS#π"β#∃β≡9β≠␈∪7W3∂#∃9↓∧{;∃β≡9β↔6+9β←O≠ 4+&CπQ↓∩YIu]r↓α7?⊗+?[↔∩aβ?;*β∂π9εCπ[∃π;'O#/→βπ␈+Qβ?v)∨Mβ7+SWK*β←'OF+M84TK9β≠∞≠Q1βO!β'Mε	β∂?w≠↔GW.s∂∃β}1βC↔␈β3∃∨~β≠3W∨#WπSNs≥β7␈#'[π&K?;Mπ##πPhS?;∃ε{≠S↔rβ←'OF+MβSzβK↔O'∪π'9ε{;∃∨~β≠WS/∪∃β←O≠#↔Mε3K?5π≠πS'≡3π∂SN{984U##WMε{;∃β⊗{S!β>;SMπ#=β∂}sSK?bβS#∃π;'O#/→βS#.kO↔36+Mβπv!β↔[.p4+O'∪?;∨/⊃βS=εKKπv;∃βSzβC#g≡K∂π3gIβ≠K/≠SKπ&)β∂↔↔#π'9ε3WSW⊗)β←'≡C↔M8hRS#∃εc'S↔⊗Keβ/Cπ7Cf)β'M∧{∪gO≡+WMβF[';:β#'Mεk↔9β&K∃β#NiβS=π##∀4VkπOQbβC3W:βS#↔O⊃β?←rβ↔πK~aβπ;"βπ∨K.)β;?"βS=β}∪↔eβFK5β←F+9β#(h+←'faβπOZβS=β⊗)β≠K.+⊃β←F+9β#*β#↔π↔→βS#*αO'K.sM84Ph(&←*β∂?Wf!β∪↔6K;∃β
βC↔K≡{9βSzβ∃β≡{;O'∨#↔;SgIβ7?&K[πS.!β'_hS#∃βFMβ;zβ∪?↔≡q∨Qβ/3↔Iβ>KO!β&yβ≠K/≠SKπ&)β?Iε≠#π;>)β#'~β≠WS/∪∃β7␈#'[π&K?;M`h+W"β';O&+π⊃β>KO#↔~βS=β≡S'O7IβS#.i9↓α6{IβO.≠!β¬πβ↔KO}q1β'"β7πdhS7π/*βO↔;≡)βS=εK;S↔?∪πS∃π+S'3O#eβ?6+Iβ#O→β3'6+S'7*q↓α#*β←'3bβ?↔Hh*↔fc7π9?→βCKNs∂'Cf)β?→ε{CS'n3'SJp4(4PJS#↔⊗)β'Mε	βO↔w≠∃β'rβ←#'≡AβC↔␈β3∃β∂∪∃[>+π/3J[∂?w≠'OS.sQ4VK9βSF+'Iβn{S'[∂#'?;~q↓α;∞k↔3ebβπ9β∞c∂?#}c'
β}3S↔9π;'O#/→βS#∂ 4+#*β←↔K*β;?Qε3∂?F{3'
bβπSS.kCSMπ#=βK.∪+W∨!β#'~β7?SO3πS'}sM1β∞sm$cd
attempts to frustrate his future desire for drink.  All this while
he is in self-improvement mode.  When he is on a drunk, he drinks,
but he doesn't wish that he were never in a self-improvement mode.
In fact the reverse, he can simultaneously want a drink and wish
that he didn't want a drink.  Thus at the higher level of what he
wishes his motivations were, he is consistent, and we can call
this weak consistency.  Presumably he can wish he wished his
motivations were different.

	Now suppose that a person can control either his future
motivations or his future physical possibilities.  To the extent
that he can, his motivations will become more consistent, or at
least his behavior will become explainable by a more consistent
motivational structure.

	Now suppose we have two or more people.  Can we establish
a motivation for this ⊗society?  In general we cannot.
In the first place, the members will have wishes about each
others motivations, and in the second they may love, like,
dislike or hate one another.  The interpersonal motivations
may be specific or general.  John may want Mary to love him
or he may want someone (anyone) to love him.
However, one can almost always establish some social preferences.
Thus there are some pairs of alternate events about which all
preferences agree.  One can often do more.  The members of
the society may be impersonally motivated at least part of the
time, i.e. they may be motivated to establish laws that govern
all equally.

Aside: The fundamental human right should be to secede from
society.  This was feasible when there was an open frontier,
and it should be a social goal to re-establish a frontier in
space.  Once this is possible again, the demands that any society
can make on an individual will be limited by his option to
secede.
Actually secessions will take place in groups, and this complicates
matters.  (The groups form for practical reasons, and because humans
are socially motivated.)

	Where social choice is necessary an approximate social
welfare function must be constructed or else unnecessary suffering
will occur.
How misunderstandings about human motivation lead to harm

	People often have unrealistic expectations about achieving
and sustaining happiness.  These lead to dissatisfaction with
institutions that are working as well as can be expected, but
cannot make people happy.  This may lead to these institutions
being replaced by worse institutions.  At least this seems to be
the dominant trend at present.

	Animals aren't happy.  Primitive man isn't happy.  There is
no reason to suppose that the capacity for continuous happiness,
whatever that might be, has been evolved.

	What can society do?

	1. Society can increase the length of human life and reduce
the number of people who die prematurely.

	2. Society can increase the options available to its members,
but it can't assure that exercising these options will make people
happy.

	3. Society can reduce the occasions for conflict among its
members.

	With more technology, more might be done.

	1. It might be possible to modify the human brain or drug it
to increase its capacity for happiness, but then we really would have
to define it first to know what we were increasing.  If we could do
that we might not want to for a variety of reasons.

	The drug culture of the 1960s was under the collective delusion
that the various drugs could increase happiness.  This turned out to
be yet another illustration of the fact that collective enthusiasms
can induce people to claim all sorts of things about themselves.

	However, it should be noted that the drug culture was based
on no scientific theory whatsoever about the actual effects of the
drugs on the brain.  It is not excluded that a scientific theory of
this kind will eventually be developed that will lead to effective
psychoactive drugs of a more acceptable kind.  It still may be a bad
idea, but if a plausible candidate is developed, enthusiasts may be
allowed to try it on themselves and we can observe them.  Other means
of personality modification may also be tried.

	2. If we understand better the tendencies for young men
to develop group hostilities - the basis of ancient tribal warfare -
we may be able to prevent it.  It would be interesting to know the
exact mechanism whereby the generalized pro-tolerance propaganda of
my school days was converted into the hostile ideologies of today.
Restoring this generalized propagand for tolerance that produced the
peaceful and prosperous 1950s would be very worthwhile.
Redesigning our motivations

	While we must face the reality of the motivational sturctures
we have evolved, we don't have to be content with them.  The religious
notion of original sin can be construed as asserting that the natural
motivational structures of humanity are unsatisfactory and advocating
that they be redesigned according to certain principles.  The only
good reason for so construing the notions is to see if something
important can be salvaged from the enormous religiously motivated
literature on morality.  Whether this is worthwhile is an open
question.

	Here are some of the ways in which we might like to redesign
human motivations.

	1. We would like to make fulfilling our long term goals a
stronger short term motivation.  This is our way of expressing a
desire for more will power.

	2. We would like people to be good.  It might be supposed
that this could only be discussed scientifically in connection with
a full definition of "good".  This is a mistake.  There is nothing
to prevent us from proceeding as though "good" represents an
approximate and incompletely known natural kind.  Thus we may suppose
that what it means for an action to be good has an approximate
objective existence even though we aren't ready to specify what it
is.  This certainly the way we teach children, and unless we want
to postpone trying to do good until a full definition is obtained,
we must take the same attitude as adults.

	We say "approximate natural kind" because we expect that
an ultimate analysis is impossible.  Thus for any notion of good,
there will be more or less weird circumstances in which the notion
gives paradoxical results and where intuition is much help either.

	There are several potential ways of modifying motivations, and
some of them seem dangerous.  However, the really dangerous ways seem
inacessible for the present.  One way is brain-washing, i.e. indoctrination
under conditions of full control of information and the ability to
subject individuals to interrogation, punishment and provocative
temptations.  The Chinese experience of the cultural revolution where
a billion people were subjected to ten years of such indoctrination
seems to prove that brain-washing is extremely ineffective.  The
ideas of the cultural revolution seem to have vanished as soon as
the compulsion weakened.

	It is certainly fortunate that Mao failed.