Presented at the

Presented at the "3rd Annual International Conference on Thinking"
Harvard University, 1984

Epistemics and Social Cognition

John G. Schmitz

Ph.D. Program In Educational Policy Studies

University of Illinois

The literature on social cognition and the psychology of

reasoning present a picture of ourselves as "intuitive scientists"

who move through the social community gathering evidence and

making hypotheses and predictions about the behavior of the

persons we encounter. But although our theorizing may indeed

resemble actual scientific methods, the comparison only goes so

far. The literature from these disciplines, highlighted by the

integrative work of Nisbett/Ross (1980), seems to indicate that

humans are susceptible to multiple biases at several stages of the

judgment process. The seriousness of these results stems from

the high frequency in which the biases affect judgments as well

as from the pervasive influence which they exert. Furthermore,

the biases of judgment seem to be mainly the result of

`cognitive' rather than `logistic' and `motivational' sources.

Therefore they are likely to be rather permanent, and not

temporary, characteristics of our thinking.

In this paper I discuss one group of the so called 'bad

habits' which the social cognition literature has uncovered: use

of the availability heuristic (Kahneman, 1973). For now I can

only express my belief that my discussion of this group of

heuristics will apply to others as well (e.g.. the anchoring

heuristic). I then discuss the discipline which Alvin Goldman

(1978a) has named "Epistemics", a "regulative theory of

cognition" which is closely tied to cognitive psychology, and its

potential for the analysis of the cognitive habits under

discussion. I also discuss the notion of "epistemically

responsible action" (Kornblith, 1983), a weakened criteria of

justification. I am concerned with the general question of what

constitutes justified (or epistemically responsible) action in the

social sphere, though I only aim at a small part of the required

answer. It should come as no surprise that I view the capacity to

attenuate the effects of deep seated biases as an important

feature of epistemically responsible action. Because I view this

capacity with such importance, I consider any precursor of it to

be extremely important as well. I consider metacognitive

knowledge just such a precursor. I argue that knowledge of the

limitations (and capacities) of one's mind offers the individual

cognizer greatly increased opportunities for offsetting the

effects of the biases, thus offering increased opportunities to

make epistemically responsible social judgments.

Metacognitive knowledge is an object of increasing importance

to researchers in the cognitive sciences. Flavell (1976) has

defined metacognition as the following: "One's knowledge

concerning one's own cognitive processes and products or anything

related to them", also as (Flavell, 1981): "Knowledge or

cognition that takes as its object, or regulates, any aspect of

any cognitive endeavor". Flavell (1981) notes the various areas

in which metacognitive knowledge has been suggested to "play an

important role" in. These areas range from oral communication,

reading comprehension, language acquisition, attention, memory,

and social cognition to other "diverse forms of self control and

self instruction." Many other researchers are working on

metacognition as well. (Those interested in pursuing the topic

are advised to see the Flavell work as well as Brown 1978).

In this paper I am mainly interested in the role which

metacognitive knowledge could play in the design of value

education program. I claim that teaching metacognitive

knowledge, in some form or another, should be an important

element of value education strategies, just as it should be and

and is becoming an important element of cognitive education

strategies: the systematic teaching of thinking skills.

The view that I give will be appreciated only by those who

favor the view that people can control their thinking, either

directly or indirectly. And it will be appreciated by only those

who also believe that cognitive biases can be avoided, or at least

lessened to some extent, by some form of control. In this

respect it should be clear that I have aligned myself with the

information processing paradigm , particularly with research which

stresses the role of executive control of cognition (cf.

Underwood, 1978, 1980; Moray, 1978). Lastly, only those who

agree that the biases are inappropriate and ought to be

avoided, will appreciate my view.

Social Cognition

A recent research trend which carries with it sweeping

implications for our understanding of human cognition concerns the

shift from largely motivational and logistic accounts of the

sources of judgmental biases to largely cognitive accounts. This

trend is marked by its emphasis on the effects which cognitive

biases bear upon stereotyping (Hamilton, 1977), impression

formation (Wyer/Carlston, 1979), and other processes of social

judgment. Characteristics of selective attention (Evans, 1983;

Kahneman, 1973; Wickens, 1983) and accessibility of stored

information from memory (Higgins & King, 1981) are examples of the

sources of particular biases.

Several cognitive biases, striking in their range of use, are

referred to under the heading of construct accessibility . The

'construct accessibility' literature has exposed several cognitive

biases which have the effect of favoring recall of certain

information in memory, as couched in constructs of various kinds

(e.g.. schemas, stereotypes, scripts), over other stored

information. This is important for the following reason. The

more accessible constructs are from memory, the more likely they

are to be employed in judgments. The more likely they are likely

to be employed, the more likely that other relevant stored

information will be ignored.

Higgins and King (l977) list four determinants of construct

accessibility: How recent a construct has been stored, how

frequently it has been recalled, how salient or vivid the

construct is, and how strongly the construct is related to

currently employed constructs. The effects of vivid information,

for example, reveals the influence of construct accessibility in

many situations. Several experiments expose the willingness of

subjects to change their opinions in the face of vivid but

statistically insignificant information, and their unwillingness

to change their opinions in the face of boring but statistically

significant information (Nisbett, et al. 1977). As we all should

recognize with a little reflection, vivid information is tempting

fare in many situations. (Advertisers know this all too well.) In

short, a variety of research has demonstrated that judgments are

often based largely upon information which has been recently

presented, or which has been often recalled, or which is highly

vivid and memorable.

`Construct accessibility' stands for an effect which is

largely if not completely equivalent to the phenomena which

Kahneman and Tversky (1973) have called the "availability

heuristic": A heuristic in which the judgments of agents are

biased by the "subjective ease of generation" of stored

information. The nature of our largely automatized recall

mechanisms seem to be responsible for maintaining the heuristic.

More than any other heuristic, availability seems to constitute a

built in bias of judgment. And it can be argued that its effects

are perhaps more pervasive than any other. Research indicates

that it occurs extremely commonly and, furthermore, largely

unconsciously (Nisbett/Ross, 1980). Since it is an effect which

precedes judgment, perfectly valid reasoning can frequently

turn out to be unsound: The information `presented' to the

reasoner was biased to begin with. An adequate demonstration of

the claim that this heuristic figures strongly in social

judgment is not possible given the lack of experiments which

`track' an agent's information processing, especially selective

processes (Evans, 1983), across the intricate terrain of everyday

social judgments. Only with such `big picture' data could we

truly demonstrate the frequency with which the availability

heuristic is employed. But there is some data which may give us

an idea of how broadly the effects of availability reach.

One source of the pervasive influence of construct

accessibility effects derives from their long lasting influences

upon judgment. Consider what Higgins (1977) has dubbed the

'sleeper' effect: Highly accessible constructs not only seem to

monopolize processing in the short term, they exert a persisting

influence upon subsequent judgments which can, and often do, grow

stronger and stronger as the construct is relied upon more and

more. Constructs can breed quite powerful structures which are

heavily resistant to 'deconstruction'. It goes like this. An

initial judgment, biased by availability or some other,

heuristic, produces a construct. The construct in turn is used

in a few subsequent judgments and then more and more judgments,

figuring to a greater or lesser degree. We can imagine a map of

the judgments involved as fanning out into a broad network from

the initial judgment. This is an idealization for sure, but it

is also highly intuitive knowledge is cumulative, structures are

built upon and are directly dependent upon preceding structures.

So the generation of reliable constructs becomes an important

matter, especially when entrenched constructs happen to be about

other persons.

One particular kind of recency effect on judgment (once

again, the tendency to bias in favor of recently stored

information), priming , provides an especially appropriate

example. Consider the following experiment (Wyer, in lecture).

Subjects are asked to form an impression about some person who

they have been given a brief description of. They are asked to

form an impression of "Mary", "a person who let someone copy

answers from her exam". Subjects can be manipulated to form

certain impressions by first priming them with constructs which

tend to generate the impression. Before the impression formation

task, subjects read or hear some information in which a

particular construct is heavily favored, for example the

construct `crime'. Subjects can be primed to form impressions

like 'dishonest', as well as `kind' and `smart', depending on the

particular construct which primed them. All impressions are

reasonable in a way. Mary can be thought as dishonest because she

cheated. But she was also smart enough to know the answers and

furthermore she was kind enough to help out some poor soul who

didn't have them. Clearly the subjects are falling prey to the

availability heuristic: just because certain constructs are highly

`available' to them, they're likely to form impressions on the

basis of the constructs.

If I may comment briefly on the 'normative appropriateness' of

the availability heuristic in this situation, the first thing

that strikes one is that the resulting impression has an extremely

arbitrary air to it: Any of the three different impressions could

have been formed, dependent only on "the subjective ease of

generation" of the particular concepts which primed the judgment.

The construct accessibility literature is a bit chilling in its

implications for the malleability of our impression formation and

other social judgment processes.

On one hand this leads us to talk about individual

responsibility to combat such effects and the normative

foundations for such responsibilities. But we could also, and

probably should, talk about broader responsibilities which

corporations and other entities should recognize. Much of the

strategy that ad campaigns use specifically exploit cognitive

processing limitations which are largely automatic the processing

of ad information is usually not a matter of choice. For instance,

it's often not a matter of choice that sex role stereotypes become

deeply entrenched in our networks of belief. So eventually a full

discussion should consider full fledged efforts, individual and

societal, to lessen the extent to which individuals can be biased

by cognitive effects like availability. Such a discussion could

go a long way in promoting the freedom of agents to truly choose

to arrive at judgments, about products and about other persons.

I now move on to an account of how the availability heuristic,

and hopefully others, can be evaluated.

Epistemics and Epistemic Responsibility

Epistemology can be considered as the ethics of belief:

what ought we and ought we not believe? All too often the

asking of this normative question has not been applied

to judgment in the social realm. We intuitively recognize that

social judgments can be better or worse, true or false, but

philosophers have not tried very hard, if at all, to give

an account of justified social judgment. Beliefs which are a

result of impression formation, attribution, stereotyping, and

other processes of social judgment have not been subjected to

serious epistemological analysis. Only recently has the question

begun to be asked and the solutions begun to be searched for.

Nisbett/Ross, for example, frame the question as concerning the

normative appropriateness of social judgment: What is the

correct way in which to arrive at a belief about another person,

or even about yourself? What are the normative criteria of social

judgment? It can be argued that the reason philosophers have

shied away from considering these questions is because

epistemology has not had the resources to handle the task. Now

there is reason to think that it does.

Alvin Goldman (1978) suggests that epistemology should

undergo a reorientation. He calls the new orientation

'Epistemics'. Epistemics represents an attempt to integrate the

powerful theoretical apparatus of cognitive psychology with the

normative apparatus which we use to evaluate judgment and belief.

Because of its psychological orientation, it constitutes a radical

departure from traditional epistemological methods. It offers the

possibility of detailed accounts about what cognitive processes

should be invoked to perform some task and how they should be

invoked. Previous epistemological work has neglected this

possibility because of its overwhelming emphasis on purely

logistic criteria of truth and justification. It has previously

been unfashionable, to say the least, to discuss psychological

conditions of justification. This new look which epistemics

offers meshes well with the new emphasis which psychological

research has made on non logistic sources of errors in reasoning.

Indeed, Goldman's awareness of the cognitive bias literature was

apparently a major impetus for his development of epistemics.

Part of the strength of this approach derives from being

anchored in our 'natural epistemic positions': Epistemics is

geared to analyze our cognitive activity and prescribe 'preferred'

kinds of it on the basis of what we can and cannot do . In this

way Goldman's project constitutes one approach of the general

movement Quine (1969) has named "Naturalized Epistemology".

Epistemics, as Goldman conceives it, is charged with the project

of improving human cognition now , hence it is concerned to

develop an "executable", not ideal, normative model for the

analysis of human cognitive activity there is little reason to

prescribe what we can never do. Furthermore, there is little

reason to prescribe what is grossly impractical to do. Thus the

focus moves away from more traditional epistemological concerns.

Highly idealized epistemic 'rules' which only could be implemented

given a digital computer and lots of time are to be avoided.

Being told to follow the rule of deductive closure is just such a

non executable rule checking one's belief system for absolute

consistency does take time to say the least.

Perhaps the main strength of epistemics is in virtue of the

precise and empirically well founded rules which it can generate,

a result of the tight alliance between epistemics and contemporary

psychological research. As Goldman notes, for example,

traditional epistemology has not been sensitive to the now common

distinction between occurrent and dispositional belief, or

between different kinds of memory storage. Without such

distinctions, prescriptions are bound to be too vague, to miss

the mark because they don't have the empirical story right. For

instance, if subject's are prescribed to believe a certain

proposition in order to be justified in holding another, is it ok

that the proposition is not stored explicitly, or can't be

remembered? What counts as holding a belief? Shortcomings in the

psychological vocabulary which traditional epistemology has

employed has contributed to distancing its results from practical

everyday implementation. In particular, `flaws' in the system,

like susceptibility to biasing effects, must be continuously

taken into account. Rather than prescribing rules which an ideal

cognizer, free of such flaws, would follow in decision making,

epistemics is an attempt to prescribe practical rules which make

the best of the processes which humans have at their disposal.

But as of yet we do not have an actual criteria of warrant to

use for this project: How exactly can social cognition be

appraised? Epistemics offers only a general methodology, an

approach which is compatible with several views about

justification, but as it stands it does not offer particular

criteria of justification. As a start in this direction, I would

like to introduce a refined version of a criterion offered by

another philosopher.

Hilary Kornblith (1983) defines epistemically responsible

action as the product of an agent " who has done all he should to

bring it about that he have true beliefs ". His view is of

interest in virtue of it's emphasis upon components of

belief formation of a largely non logical nature. Because of this

emphasis, justified belief (or "normatively appropriate" belief)

is more than just a function of the reasoning used to reach

beliefs, but also a function of how selective processing (the

stages involved in the gathering and storing of information) take

place. Kornblith argues that to merely instantiate appropriate

logical relations between beliefs (which is the criterion of

ideal reasoning accounts of justification) fails to catch defects

in the manner in which people gather information about the

theories and beliefs they come to hold. The point is that

gathering evidence can easily become a self fulfilling prophecy in

the absence of adequate selection, schematizing and storage of

information what you want to see and hear you will. So it seems

important that people responsibly select information for

judgments. Yet it is hard to see how an ideal reasoning account

can handle such concerns.

The role of epistemic responsibility can now be made more

clear. Detailed psychological appraisal of an agent's

belief formation processes has been a task resistant to tradition

epistemological analysis. Another criterion which supplements

traditional methods of appraisal must be added. Epistemic

responsibility is a good candidate. Part of it's job is to

enforce minimal conditions of justification: certain things must

be done if a belief is to be considered epistemically responsible.

The things that should be done are directly tied to the capacities

and limitations all humans possess. There are a variety of

cognitive endeavors which, given the nature of our cognitive

systems, cannot be justified. Beliefs formed by hastily scanning

a distant object go against the limitations of the visual

information processing system. Beliefs formed in periods of

divided attention are another example. Beliefs formed from faulty

memories are another (examples from Goldman, 1978).

So my 'epistemic analysis' just comes to this. There are

minimal conditions of responsible social judgment, though as of

yet unarticulated and unformalized. These conditions can, in

part, be derived from the very nature of our cognitive systems.

Priming , for example, seems to be a fact about our the way we

gather information: it's just a fact that we are naturally

susceptible to recently presented information. From this fact we

can then derive a general condition which in some way promotes

resistance to priming effects where appropriate. So I have in

mind a cluster of basic conditions the realization of which

constitutes epistemically responsible judgment. The particular

condition which I offer concerns self awareness, to the extent to

which it is possible, of the occurrence of biasing effects during

decision procedures: It is epistemically responsible to make

yourself aware that certain products of your cognitive processes

have been generated through the use of heuristics.

The derivation of general conditions of responsible social

judgment is not straightforward. Multiple barriers face any such

attempt. Here's one sort of objection: A variety of

different and perhaps contradictory conditions could be generated

from the same 'fact' about our minds. One way to make the

objection more concrete is this: the particular context (social,

physical and personal) in which a priming event occurs will always

yield a unique condition to follow. Hence, no generalizable

conditions are possible, or even desirable. An awareness

condition may be appropriate for one sort of priming event but not

for another. Also, it could be argued that it is unclear whether

there is any sense to the expression 'same fact' to begin with.

What really is a priming event? When is an event and its context

sufficiently the same to be covered by the same general condition?

One response is that research has sufficiently picked out

regularities of social judgment which, although not

context invariant, are sufficiently common to allow a general

analysis and the generation of a general condition. The condition

won't cover all cases but it will cover most. And certainly more

fine grained descriptions and conditions will always be sought

after (which will probably follow on the heels of advances

in psychology). Also, the individuation of cognitive events

will indeed be sloppy , but that will create problems only

for marginal cases, not cases in which it is relatively assured

that you are analyzing a priming event. And it appears that

marginal cases will be the exception, not the rule. Problems of

individuation will begin to disappear in the wake of a more

powerful psychology as well.

These responses are only tentative but promising ways to block

the objections I noted. I regret that I cannot pursue these

problems and their unfortunately numerous relatives, much more

deeply (which I am doing in a work in progress). Now I will move

on to a more particular application of the conditions which I have

in mind. Namely, what does this approach which I am promoting

recommend regarding the use of heuristics?

In the same way that deductive closure is a non executable

rule, many of the formal statistical strategies considered to be

normatively appropriate, are, if not quite non executable,

highly implausible as everyday decision procedures. Besides

outright non executable rules, there are highly time consuming,

and resource hogging rules. Because these are constantly faced

`boundaries', our criteria of epistemic responsibility must allow

the use of heuristics. It must gear its analysis of heuristics to

our extreme need to employ them and to employ them often. Given

the nature of the complex information the social world presents us

with, any view which advocates that formal decision procedures

are the only 'legitimate' means of arriving at social judgments,

is a view which has removed itself from our cognitive and social

realities. We must use heuristics, indeed, we may not have any

choice. But this is not to entirely condone their shortcomings as

social judgment procedures. `Shortcuts of reasoning' often lead

us to the darker side of human cognition.

It seems clear that we can 'attenuate' the effects of the

heuristics. The criterion of epistemic responsibility should

enforce the use of attenuation strategies by agents. Epistemic

responsibility should be used to suggest rules which promote

cautious employment of heuristics. Of course, it is a non trivial

task to spell out what cautious use of heuristics amounts to,

especially since we are leaving the safe confines of formal

appraisal of belief. But it is a task which cannot be avoided if

we are to arrive at a suitable way of evaluating social judgment.

What is it to use heuristics cautiously? It is to be

epistemically responsible when you use them. One part of

realizing that (and only one part) is to have some degree of

awareness that heuristics are being used. I do not mean direct

awareness of the early selective processes themselves. They occur

too quickly to be directly made aware of. But we can be aware of

their products: the beliefs which result from them. Given such

awareness of the products of the early processes, along with the

knowledge that they are products of such and such early processes

with such and such limitations, we are in position to control any

attenuation of the effects of the processes.

Anyone who promotes an awareness condition, like I am doing,

has a lot of work cut out for them. Here are some examples.

First, it seems that an eventual goal would be for a person

to automatize (Schneider/Schriffrin, 1977) procedures for checking

upon the products of early processes. Surely it would be the

most effective and resource efficient route to take. And if this

is the case, the awareness condition which I am offering has a

very unclear future.

On one hand, this is fine. I could promote the condition as

a short term rule which can be eventually discarded. This is to

claim that active control is necessary for self monitoring at

first, but later becomes dispensable. So conditions must evolve

along with success in automatizing checking procedures. All the

same, I think the objection can and should be blocked for other

reasons.

My response centers on the apparent novelty of the

circumstances of judgment, social and otherwise. Circumstances

are such that novel responses, which seem to require novel

control of responses, may be necessary. Perhaps many responses

cannot be 'programmed' into a person without loss of the ability

to self regulate in situations where odd stimuli present

themselves. I recognize that many common responses can be

successfully inculcated, but I claim that novel control will be

necessary in many other circumstances. If so, the awareness

condition will endure as long as the need for novel control

endures.

Second, there is some very serious work cut out for

`awareness' advocates in the form of replies to the body of

literature surrounding the seminal paper of Nisbett and Wilson

(1977). What is the reliability of our access to the contents of

our minds? What exactly do we have access to? To what extent are

we merely mouthing conventionalized responses when we report on

why we did so and so? Or why we think so and so of some person?

And to what extent are we perpetually overconfident in reporting

our motives for acting, or for attributing motives in others? And

are we so controlled by heuristics that the self attenuation of

them becomes very unlikely?

Indeed, the problems seem so pressing and numerous that even

rejection of a few of them will still leave many. In fact, after

a reading of Nisbett/Wilson (1977) or Nisbett/Ross (1980), I think

that many people would regard my claims as grossly implausible.

How could persons with that kind and reliability of introspective

access have any hope of self regulating themselves?

I should start out by noting that I generally think that the

Nisbett/Wilson claims are: A. considerably over generalized, and

B. dependent upon experiments which are easily interpreted in

several different ways. I cannot cover these claims in this

paper, though I will develop one of them to a small degree. (I

am currently writing a thesis length discussion of metacognitive

strategies in the light of Nisbett/Wilson type objections).

I think that the 'permanence' of the problems which

NIsbett/Wilson have pointed out concerning cognitive processing

about the self have been exaggerated. They present results which

show that subjects report information that they could not have

access to. Apparently, in the experiments they ran, the subjects

substituted a fabricated causal story for the 'true' explanation

behind a piece of their behavior. They could not know the answer

they gave yet they gave it anyway they told more than they could

know.

But Nisbett/Wilson are uncareful in their conclusion. For

instance, they do not consider just what the effects of more

reliable 'causal theories', produced by possessing metacognitive

knowledge, might be on the tendency for agents to give 'mixed up'

accounts of their behavior. This could be tested. Experiments

similar to the ones which Nisbett/Wilson cite could include a

'metacognitive condition', where subjects have been trained with

metacognitive knowledge, in order to compare `introspective

performance' with a group which didn't have it. Some may claim

that this has already been done. I don't think it has. The

success of metacognitive strategies in increasing the reliability

of introspective reports will require systematic training of

individuals with metacognitive knowledge, which hasn't been done.

Piecemeal training programs don't go very far.

In general I think that the Nisbett/Wilson claims can be

watered down quite a bit. If this is the case, a major barrier

to the feasibility of metacognitive training for a variety of

applications, including the attenuation of the effects of

heuristics, is removed.

The upshot of much of the social cognition literature, for my

purposes, comes to this: `Early processing' is important.

Cognitive processing at the selective levels are the source of

biasing effects which exert a powerful and lasting effect upon

subsequent social judgment. The limited range of selective

attention, the limited capacity of short term memory, and the

very fact that we process things in stages, many of which are

unconscious and automatic, are among the reasons which make this

so. Because of the crucial effects which early stages of

cognitive processing have upon the judgment process, and because

agents do not often 'go back' to check those effects, I claim that

metacognitive awareness of the nature and occurrence of early

processing is essential to epistemically responsible social

judgment.

Being epistemically responsible does not require us to engage

in long, laborious and often trivial decision making just to form

an impression, but it does require that we, in appropriate

situations, have some degree of awareness of what has happened. If

we form an impression on the basis of highly recent information,

so be it but knowing that we did so is essential to the manner in

which we eventually weight the impression or categorization or

social judgment of any kind. How we weight the value of the

decisions we reach I want to remain neutral on, but that we

should be in a position to is, I hold, an important condition of

epistemically responsible judgment.

Value Education

Value education strategies have proliferated. At least

five different general types of strategies have been suggested to

span the field: inculcation, moral development, analysis, values

clarification, and action learning. Several others can be

imagined merely be combining the basic ones. I am attracted to

components of several of them, probably formed around a core

analysis approach. However, in the case of few of the strategies

which I have seen, have they seemed sufficiently cognitively

oriented, that is, geared to teaching cognitive skills for our

thinking about others. None appear to be sufficiently

metacognitively oriented. In addition they do not appear to

explicitly and systemically handle cognitive biases of judgment,

though they are often intended to offset their effects.

Because of these limitations, I claim that such approaches

are too superficial to be of strong and lasting benefit. It seems

that this claim can be advanced against many contemporary value

education strategies. Moral reasoning seems to involve much of

our complement of cognitive skills. If we agree with the claim

that metacognition is an important component of cognitive

education (again, the systematic teaching of thinking skills),

then the same follows for value education instruction.

Furthermore, there is a even greater imperative for this project

then there is if we were only interested in teaching, for

instance, problem solving in the sciences.

The imperative I have in mind is moral imperative: the

responsibilities we have towards others. It can be argued that

belief forming processes must be appraised differently when their

'products' are about people than when they are about rocks and

trees and stars. Social judgment, unlike judgments about

material objects and events, brings the judgment under the

appraisal of the moral community. Epistemic appraisal covers both

sides of the fence on both sides there are cognitive processes

and their products, beliefs, to be appraised. But on one side

the beliefs are about persons which we have responsibilities

towards. On this `side', then, there is an heightened interest to

evaluate belief production as well as to suggest ways in which

such production can be enhanced. The force behind an account of

epistemic responsibility is greatest in the social realm simply

because it is there that it matters most. We all know what is at

stake when impressions are formed and attributions made. So a

suitable theory of morality, which I am only imagining and not

providing in any way, provides the real impetus behind the

development and adoption of epistemic prescriptions, of the sort

I have been speculating about, regarding social judgment.

This may be an obvious point, but I don't think it everyone

recognizes it. For example, Nisbett and Ross (1980) discuss the

cost/benefit analysis of the use of heuristics. When they do,

they do not adequately consider the possibility that

responsibilities toward others should be fit into the analysis.

Whether or not it's costly to me that I spend some extra time to

carefully form an impression of someone, there are many

circumstances where it seems I should in order to be fair to them.

I suggest that epistemically responsibility is a suitable

first step for assessing fairness of social judgment. I would

hope that we had a stronger criterion, one which actually assessed

the way, for example, that we weighted information in

judgments. But that will be the result of much further work on

the part of many people, in both epistemology and ethics.

I am arguing that the evidence points in favor of selective

biases as playing an important role in social judgment.

Reflecting upon the simple fact that evidence gathering and

storage and recall of evidence from memory precedes judgment,

should remind us that often the best reasoning cannot escape the

effects of early biasing. This is, in a way, only to repeat

what we commonly teach in introductory logic: distinguish

'validity' from 'soundness'. So I am pointing out that 'sound'

social judgment is often a matter of what you start with, namely

the evidence you have particularly a matter of how you gathered

it, how you 'schematized' and organized it and how you recalled

it. To be sure, one 'reasons' that one's evidence is poor. So

doesn't this just lead us back to traditional reasoning

instruction? Yes and no. Reasoning about your judgment process

is impossible if you don't what your judgment process is . Yet

reasoning is still obviously important. My claim can now be

expressed more precisely: Metacognitive awareness fosters

reasoning which exposes cognitive biases.

The suggestion I am making comes down to this. Value

education programmes cannot get very far independent of a more

systematic and sophisticated attempt to teach students about how

their minds work. It seems to me that the problem is not so much

that we can't teach minimal guidelines for weighing evidence

properly, or teach the responsibilities which motivate students to

do so. That's what several approaches do very well now. What

they don't do is to drive home an understanding and awareness of

the cognitive processes involved in social judgment, and hence

they don't foster the sort of practical awareness that is

necessary to regulate everyday judgment.

Conclusion

In the course of our everyday thinking it seems that we

'drift' along with the flow of thought much more often than we

make an effort to reflect on it and evaluate it. But more effort

by itself isn't enough to change this. It can be argued that the

main reason we 'drift' along is that we lack an adequate framework

for ordering our thinking about our thinking. Certainly there is

a framework, what Dennett (1978) calls our 'folk psychology', but

it's weaknesses are many. Importantly, this framework, described

by one writer to be "ancient and ramshackle" (Churchland, 1980),

lacks the information which can help consistently expose biases at

the selective level of judgment. We have lacked an adequate

framework from which to monitor and evaluate our cognition.

Cognitive biases escape our attention.

It seems that we literally are coaxed into biasing our

judgments towards information which is vivid or which has been

recently or frequently stored. Automatic processing seems to be

the culprit. Information is thrust in front of our consciousness

very quickly, largely without our control. We scan information

automatically, we store it automatically, we recall it

automatically, and we often infer automatically. Only with the

effort and knowledge to reflect and evaluate upon highly

automatized thought processes are biases and their effects likely

to be exposed. We must learn to appreciate the intricacies and

complexities of the cognitive terrain which we daily traverse but

of which we are almost completely unaware. The Socratic maxim,

"know thyself", still provides us with an important message to

reflect upon.

Acknowledgements: F. Schmitt and C. Anderson.