self-efficacy

[From Jeff Vancouver (2001.08.01 13:35 EST)]

CSG conference followup:

For those at this year's CSG conference, you might have noticed an exchange
between Bill P. and me regarding something I said he said. I was relatively
flip in my answer, which I want to address more substantially.
Specifically, the issue was about self-efficacy and a comment Bill made to
A. Bandura in the American Psychologist. What I paraphrased that he said in
my talk was the following:

The Paradox:
Powers argued that when individuals have high self-efficacy (as compared to
when they have low self-efficacy) they may:
have overly optimistic assessments of the degree to which they are meeting
their goals
hence, lower discrepancies,
hence, apply fewer resources to meeting goals.
Hence, self-efficacy would negative relate to subsequent performance.

This was the phrasing that Bill objected to. What I said in response is
that it was too late, as the paper was already in press that represented the
study that was the topic of the presentation (it should be out this month
actually). I also knew that I had not exactly represented him in the
presentation and had tried to be more careful in the published manuscript
(also I had more time to get the point across). Below is what I said Bill
said. As you can see, I was careful to quote him. I would be interested in
knowing if Bill thought the quotes were out of context or my rewording
(following each quote) is a great distortion of the meaning of the quote.

"The paradox relates to the role of expectations or self-efficacy beliefs,
goal levels and performance, and was revealed in a debate between Powers
(1991) and Bandura (1989). Specifically, Powers argued that Bandura
described several variations of self-efficacy beliefs. One variation is the
�belief that one is justified in setting high goals [that, because
discrepancies from goals drives behavior] would lead to an increase of
effort and, if the effort were successful, an increase of the perceived and
actual effectiveness of action� (Powers, 1991, p. 152). In other words,
self-efficacy beliefs promote the adoption of high standards and the high
standards result in higher performance if the person actually has the
capacity to perform at that level. However, another variation is the belief
�about one�s actual effectiveness in achieving a given goal� (Powers, 1991,
p. 152). Powers argued �a belief that works by increasing the optimism in
perceptions � by representing the same actual consequences of behavior more
favorably � reduces the apparent shortfall that is driving behavior, and so
decreases behavior� (p. 152). That is, when individuals believe that they
are meeting their goals, they are less likely to allocate resources (i.e.,
time and effort) toward those goals as compared to when they believe they
are not meeting their goals. "

I am leaving for another week, so if I do not respond to responses to this
post, it is because I am not around. No other meaning should be inferred.

Later,

Jeff

[From Bill Powers (2001.07.02.1007 MDT)]

Jeff Vancouver (2001.08.01 13:35 EST)

For those at this year's CSG conference, you might have noticed an exchange
between Bill P. and me regarding something I said he said. ... What I

paraphrased that he said in my talk was the following:

The Paradox:
Powers argued that when individuals have high self-efficacy (as compared to
when they have low self-efficacy) they may:
have overly optimistic assessments of the degree to which they are meeting
their goals
hence, lower discrepancies,
hence, apply fewer resources to meeting goals.
Hence, self-efficacy would negative relate to subsequent performance.

I did say what you paraphrased, but not with the meaning that the above
might be understood to convey. What I appear to be saying is that people
who judge their own abilities highly will perform worse than people with a
lower sense of self-efficacy. That, of course, is nonsense. In the more
extended quotation following the above, you make it clear that the
prediction of a negative relationship applies only when the person is
overoptimistic about judging how close the achievement is to the goal
(whether the goal be set high or low).

The problem is that "self-efficacy" can be used in two different ways. One
way indicates a judgement about one's abilities in general. Thus a person
with a high sense of efficancy may set higher goals than a person who
doubts his or her ability to achieve well. A trained athlete will quite
realistically set the target speed for running a mile higher than a weekend
jogger, also quite realistically, would set it. We would expect the trained
runner to set higher goals, and achieve them, than a casual runner would
set and could achieve.

Whichever end of the spectrum we're talking about, we also have the case
where a judgement of self-efficacy is applied not to the goal, but to the
achievement, the performance. If a person believes he or she is achieving
more than is actually the case, the effort to perform better will be
lessened relative to a person whose judgement is that his or her
achievement is not adequate. If the goal is to run a mile, and you perceive
that you have already run that far, you'll slow down or even stop. If your
judgement about how far you've run is accurate, you've achieved the goal.
But if you think you've run a mile when you've run only half a mile, your
ultimate achievement will be a lot less. The other person, of course,
perceiving the distance already run less optimistically, will keep running
and win the race.

Best.

Bill P.

[from Jeff Vancouver (2001.08.08.0933 EST)]

[From Bill Powers (2001.07.02.1007 MDT)]

The problem is that "self-efficacy" can be used in two
different ways.
Thus a person
with a high sense of efficancy may set higher goals than a person who
doubts his or her ability to achieve well. A trained athlete
will quite
realistically set the target speed for running a mile higher
than a weekend
jogger, also quite realistically, would set it. We would
expect the trained
runner to set higher goals, and achieve them, than a casual
runner would
set and could achieve.

Bill,

The issue of the "two different ways" is the essential point of the studies
I presented. Specifically, when using a between-person approach, one will
get a positive correlation between self-efficacy and performance because
self-efficacy reflects a reasonably good prediction of performance (based
largely on past performance). This positive correlation was found and
reported in my study, along with a within-person analysis that implicated
the role of past performance in the positive relationship. The negative
relation can be found by looking at the relationship between self-efficacy
and subsequent performance within a person over time (provided they are not
demonstrating a learning or fatigue effect). It is a way of getting at the
mistaken belief that one has reached the level they sought (or the mistaken
belief arising from an overly pessimitic belief that they have not reached
the level they sought when they in fact had).

I did make on inference about your intended meaning in the paper. I
suggested that you would not suggest that this self-efficacy effect occurs
unless there is some ambiquity in the input quantity (i.e., it does not
matter how confident I am if there are clearly marked mile markers).

I could send a copy of the papers (there are now two on this topic) if you
are interested.

Jeff

[From Bill Powers (2001.08.08.1413 MDT)]

Jeff Vancouver (2001.08.08.0933 EST)--

The issue of the "two different ways" is the essential point of the studies
I presented. Specifically, when using a between-person approach, one will
get a positive correlation between self-efficacy and performance because
self-efficacy reflects a reasonably good prediction of performance (based
largely on past performance). This positive correlation was found and
reported in my study, along with a within-person analysis that implicated
the role of past performance in the positive relationship. The negative
relation can be found by looking at the relationship between self-efficacy
and subsequent performance within a person over time (provided they are not
demonstrating a learning or fatigue effect). It is a way of getting at the
mistaken belief that one has reached the level they sought (or the mistaken
belief arising from an overly pessimitic belief that they have not reached
the level they sought when they in fact had).

I wish you would rename the variables to be "high estimates of future
performance" and "overestimates of present performance." These are two
different variables, and calling them both "self-efficacy" is unwarranted.
To say they are the same variable is to say that the same factor that leads
to high estimates for future performance also leads to overestimates of
present performance. Your paper shows that this is not true (or not true
very much of the time), doesn't it? Maybe I've missed the point.

Best,

Bill P.

[from Jeff Vancouver (2001.08.09.0900 EST)]

[From Bill Powers (2001.08.08.1413 MDT)]

I wish you would rename the variables to be "high estimates of future
performance" and "overestimates of present performance." These are two
different variables, and calling them both "self-efficacy" is
unwarranted.
To say they are the same variable is to say that the same
factor that leads
to high estimates for future performance also leads to
overestimates of
present performance. Your paper shows that this is not true
(or not true
very much of the time), doesn't it? Maybe I've missed the point.

I think you missed the point. The connection is that an _estimate_ is being
made. It would be unparsimonous to suggest that the method and hence the
result of estimation is different (ie., 2 different variables). The
question is how (or whether) that estimate affects behavior. I'm arguing,
like I believe you did, that the estimate is used to determine the adoption
of references and reference levels as means to achieve higher-level
perceptions (during a thinking mode process), and, when feedback is
ambiguous, it is used to infer a current state of a variable (i.e., it
partially determines the perception in behaving mode). These two processes
have different effects on performance, but they arise from the same model
parameter.

Also, the issue of the accuracy of the estimate (e.g., overestimate) is not
particularly relevant at this stage. For example, if I overestimate _a
lot_, I will likely adopt higher reference levels and, given a particular
reference level, likely mistakingly think I reached that level more readily
than if I _over_estimate _a little_. In other words, the sign of the
miscalibration is not the issue; the co-variation of the estimations with
performance is the issue. To separate the different processes, co-variation
can be assessed across others or via comparison with others, versus
comparison with self over time, depending on which process one wants to
confirm (of course there are other ways of trying to assess the accuracy of
these hypotheses about the processes).

Jeff

[From Bill Powers (2001.08.09.0920 MDT)]

Jeff Vancouver (2001.08.09.0900 EST)--

I think you missed the point. The connection is that an _estimate_ is being
made. It would be unparsimonous to suggest that the method and hence the
result of estimation is different (ie., 2 different variables).

In one case the "estimate" is a prediction; in the other it is a judgement
of what one perceives right now, involving no prediction.

The
question is how (or whether) that estimate affects behavior. I'm arguing,
like I believe you did, that the estimate is used to determine the adoption
of references and reference levels as means to achieve higher-level
perceptions (during a thinking mode process), and, when feedback is
ambiguous, it is used to infer a current state of a variable (i.e., it
partially determines the perception in behaving mode). These two processes
have different effects on performance, but they arise from the same model
parameter.

Is a reference signal the same parameter as a perceptual signal? The
reference signal is determined by a higher system; the perceptual signal
comes from the input function of the system receiving the reference signal
from above. The reference signal that specifies desired achievement can be
set high or low. For either setting, the perceptual signal that reports
actual achievement can exaggerate the actual achievement, report it
accurately, or understate it. One can use control theory to predict the
outcome in each of the six implied combinations of circumstances. Of course
in reality there are many more than six possible circumstances.

Also, the issue of the accuracy of the estimate (e.g., overestimate) is not
particularly relevant at this stage. For example, if I overestimate _a
lot_, I will likely adopt higher reference levels and, given a particular
reference level, likely mistakingly think I reached that level more readily
than if I _over_estimate _a little_. In other words, the sign of the
miscalibration is not the issue; the co-variation of the estimations with
performance is the issue.

Have it your way, but count me out.

Best,

Bill P.

[from Jeff Vancouver (2000.08.09.1313 EST)]

[From Bill Powers (2001.08.09.0920 MDT)]

Jeff Vancouver (2001.08.09.0900 EST)--

>I think you missed the point. The connection is that an
_estimate_ is being
>made. It would be unparsimonous to suggest that the method
and hence the
>result of estimation is different (ie., 2 different variables).

In one case the "estimate" is a prediction; in the other it
is a judgement
of what one perceives right now, involving no prediction.

The latter is not a "judgment of what one perceives," it is what one
perceives (although it might be considered a judgment of a controlled
variable). However, it is a perception that derives, at least in part, by
predicting the effects of actions (or the effect of the perceptions of the
lower-level systems), which is only necessary when a perception of the
environmental variable is not directly available. But this is a futile
argument about words (or more specifically, the word prediction).

>The
>question is how (or whether) that estimate affects behavior.
I'm arguing,
>like I believe you did, that the estimate is used to
determine the adoption
>of references and reference levels as means to achieve higher-level
>perceptions (during a thinking mode process), and, when feedback is
>ambiguous, it is used to infer a current state of a variable
(i.e., it
>partially determines the perception in behaving mode).
These two processes
>have different effects on performance, but they arise from
the same model
>parameter.

Is a reference signal the same parameter as a perceptual signal? The
reference signal is determined by a higher system; the
perceptual signal
comes from the input function of the system receiving the
reference signal
from above. The reference signal that specifies desired
achievement can be
set high or low. For either setting, the perceptual signal
that reports
actual achievement can exaggerate the actual achievement, report it
accurately, or understate it. One can use control theory to
predict the
outcome in each of the six implied combinations of
circumstances. Of course
in reality there are many more than six possible circumstances.

The answer to your question is no, but neither parameter is the one to which
I am referring. Indeed, I fail to see your point? I see no
incompatibilities between our statements. I think what is missing from your
statement is that the reference signal level may be partially determined by
this parameter that effects the perceptual signal. For example, suppose
someone says they will give me $20 if I run a mile. I may, based on my
beliefs about the shape I am in, accept or reject the task of running a
mile. I am more likely to accept it if I believe I am in good shape. Now
suppose that the course is not well marked; that I have to estimate when I
have run the mile. I may use, as part of the information that I have run a
mile, the degree of exhaustion I feel. The worse shape I believe that I am
in, the more exhaustion I might expect to experience to indicate a mile of
running (I might say to myself "I do not know if I have run a mile, but it
sure feels like I have run a mile"). Hence, the same belief about my
capability of running the mile (and more specifically, the level of
exhaustion I would expect to experience) would go into my accepting of the
reference level (run a mile) and my perception of my current state (I have
run a mile). Okay?

Have it your way, but count me out.

What does count me out mean? Does it mean I should not cite you? Don't you
complain about the lack of citing by psychologists?

I think the thing for me to do to further this conversation is to model what
I am suggesting. That is the next agenda item for this line of research.
It will be awhile. This is complicated modeling with sticky measurement and
design issues, which I need to resolve if I expect anyone to buy my attempts
at empirical varification of the model.

If in the meantime you want to model what you suggested to Bandura in the
1991 paper, by all means, be my guest. I have found that this line of
research gets a lot of attention. It demonstrates a significant distinction
between CT explanations of behavior and cognitive ones like social cognitive
theory.

Later,

Jeff

P.S. I have not had a chance to pull down your demo from the conference. I
need to get some things off my plate first. I am hoping you will have
proven me wrong. If so, you have an exciting demo.

[From Bill Powers (2001.08.09.1208 MDT)]

Jeff Vancouver (2000.08.09.1313 EST)--

Have it your way, but count me out.

What does count me out mean?

It means that this random sort of ad-hoc theorizing doesn't interest me and
I'm not going to aid and abet it.

Does it mean I should not cite you?

You have the right to cite anyone's published work. But you're not using my
ideas anyway, so why bother?

Best,

Bill P.