PSY 6450 Unit 6 Feedback 1

advertisement
PSY 6450 Unit 6
Feedback
1
Schedule


Today - lecture U6 Feedback
In-class exercise (8 points) Wednesday




Analysis of feedback: reinforcement or not?
Exam (27 points) on Monday, 10/29
ME1 over Units 1-4 on Wednesday,
10/31
Last day to withdraw without academic
penalty: Monday, 11/5
2
Exercise

Komaki vs. Locke


In one of Komaki’s study she used feedback alone
to increase safety and maintained that the feedback
was reinforcement because behaviors increased
dramatically immediately after she implemented the
feedback
Locke contested this interpretation saying feedback
could not be considered a reinforcer because:


Feedback can never be considered a reinforcer because it
does not always increase performance
In this particular study, feedback could not be functioning
as a reinforcer because performance improved
immediately when it was introduced, rather than gradually
as is typical when behavior is reinforced
3
Exercise, cont.

For each of those statements indicate whether


You agree or disagree with Locke’s statement
Explain/discuss, using as many reasons as you can





You can bullet your points, then provide a brief explanation
under the bullet
Use material from this unit and outside material if you want
to (which will have to be referenced, APA style)
Not a long paper - 1-2 pages should do it
This is not an opinion paper - you should at
least refer to relevant material that we have
covered in this class
Questions?
4
Intro to Peterson

Both Peterson and Balcazar et al. make
the same major points:



Feedback is NOT a principle of behavior
Yet we often talk about it as if it were
Peterson

The term “feedback” has become
professional slang
5
SO1: Why is it inappropriate to talk in the
abstract about what function fbk serves?



When people address the behavioral function of
feedback generally or in the “abstract,” they usually say
that it works because it is an SD or Sr, which Peterson
says in incorrect
Feedback, information about past performance can
potentially serve any number of behavioral functions
depending upon the situation
It is first and foremost a physical stimulus and only that.
Thus, it can have any or all of the possible effects of any
stimulus in a particular situation





Conditioned reinforcer
Conditioned punisher
SD
CS (in a respondent relation)
MO
6
SO2: Why can’t feedback be either
an Sr or SD in most settings?


Peterson also states that feedback
cannot be either an Sr or an SD in most
settings
He gives two reasons for each, but deals
with them in a different order than I am
going to


First, why can’t feedback be a reinforcer?
Second, why can’t feedback be an SD?
7
SO2, cont: Feedback/reinforcer

A reinforcer must immediately follow behavior.
In most applied settings, feedback is too
delayed.
Diagram will be provided in lecture
(straightforward, the next one often causes problems)
8
SO2, cont: Feedback/reinforcer

A reinforcer must be contingent upon performance
(adventitious Sr aside). Feedback is not usually
contingent upon performance.
Reinforcement
Diagrams will be provided in lecture
Feedback
Diagrams will be provided in lecture
In most situations, you get feedback whether or not you emit
the appropriate target response, hence the feedback is not
contingent upon performance and cannot be reinforcement
(contingent? If-then relationship; but I’ll come back to this in SO 3, to see whether this analysis is correct)
9
SO2, cont: Feedback/SD

An SD must evoke a behavior within 60 sec.
Feedback typically does not.


But first, this may be a new concept for you
Operational vs. functional definition of an SD
 Operational (Malottian): An SD is a stimulus in
the presence of which a response is reinforced or
punished
 Functional (Michael): An SD must evoke a
response within 60 sec given a prior history of
SD/Sdelta training
10
SO2, cont: Feedback/SD

Operational (Malottian): An SD is a stimulus in
the presence of which a response is reinforced
or punished.

Why operational?
It tells you what you need to do (what “operation”
you need you need to perform) in order to make
something into an SD, but it doesn’t tell you the
effect or function that the stimulus must have in
order to be an SD

In other words, you could perform the “operation” but for
various reasons the stimulus might not come to evoke a
response, and if it doesn’t evoke a response, then it isn’t an
SD
(x rays, high frequency tone, visual stimulus for a visually impaired individual. We have functional definitions
for consequences why not for the SD? Malott would agree that a stimulus is not an SD unless it evokes a behavior)
11
SO2, cont: Feedback/SD

Functional (Michael): An SD must evoke a
response within 60 sec given a prior history of
SD/Sdelta training

Why functional?
It tells you what effect the stimulus must have on
behavior in order to be an SD

That is, it must evoke a behavior within 60 seconds after it
is presented
(x rays, high frequency tone, visual stimulus for a visually impaired individual. We have functional definitions
for consequences why not for the SD? Malott would agree that a stimulus is not an SD unless it evokes a behavior)
12
SO2, cont. Feedback/SD
SD
Diagram will be provided in lecture
Feedback
Diagram will be provided in lecture
Be careful! It is not the delay between the response and consequence
that is relevant here (as it was for the argument about why feedback
usually cannot be a reinforcer). Rather, it is the delay between feedback
as an antecedent stimulus and the response!
Diagram will be provided in lecture
13
SO2, cont: Feedback/SD

For an SD, the response must be reinforced in the
presence of that stimulus but not in its absence. With
feedback there isn’t usually an S∆.
SD
Diagrams will be provided in lecture
Feedback
Diagrams will be provided in lecture
(for exam, not sufficient just to say, no S∆, explain; just like Malott’s procrastination eg with rgb)
14
SO3: Is Peterson right in saying feedback
is not contingent on performance?
Feedback is not contingent on performance:
R (decrease electric use)
–––> Feedback
No R (do not decrease use) –––> Feedback
But is it?
Feedback is not a unitary stimulus. Doesn’t feedback
differ depending upon whether performance is good or bad?
R (decrease electric use)
––> Feedback
No R (no decrease/increase) ––> Feedback
Feedback
graphs will
be added in
lecture
Are those two feedback graphs the same stimulus?
So, isn’t feedback contingent upon performance after all?
(one of the reasons why feedback can’t be a reinforcer; click, then click again)
15
SO7: Why is it that if reinforcement already
exists, fdb might improve performance further?

Feedback may have discriminative control and evoke higher levels
of performance due to generalization of past reinforcement
contingencies
That is, in the past, when feedback has been introduced, you have been
reinforced for performing well and criticized/punished for performing poorly.
Thus, in the current situation, as long as performance isn’t maxed out, it
may evoke higher levels of performance as an antecedent stimulus
(stimulus generalization)

Once the worker is performing better due to the feedback, the
higher performance results in more rewards (due to the already
existing differential reinforcement system), and the greater rewards
then sustain the higher performance over time
(skipping SOs 4,5,&6, straightforward but fair game; already a differential reinforcement system in place but no fdb If you add feedback why might performance increase further - assuming, of course, it hasn’t maxed out: two parts!)
16
SO7: Example


Assume that workers are being paid a per piece
incentive for each widget they assemble. The more
widgets they produce, the more money they make. But,
they are not getting feedback
Now, weekly feedback is introduced


Feedback increases the number of widgets produced because
of generalization of past contingencies (stimulus
generalization, SD)
They earn more money now because they are producing more
widgets than before, and the extra earnings now sustain their
higher levels of performance
(both points are important for the analysis!)
17
SO7: What evidence do we have that feedback
may enhance the effectiveness of extant rewards?

Bucklin et al. (JOBM, 2003, 2/3, pp. 6494)


Annecdotal from Bill Abernathy’s work:
Monetary incentives and monthly
feedback


ABA with A=incentives, B=incentives plus
feedback
“Need PM to bridge the gap”
Stongest may by Union National Bank
case study - never published
18
Union National Bank, Little Rock, AR



When: early to mid 1980s
Who: Abernathy, McNally, McAdams, & Dierks
Job: Proof operator at bank


DV: number of items entered per machine hour


Get checks and deposit slips from all bank branches and,
using a proofing machine, put the numbers on the bottom so
they can be automatically entered into the bank’s computer
Industry standard: 1,000 items per hour
Phases


Baseline
Weekly graphed feedback
19
UNB cont.

Phases, cont.

Incentive 1




Incentive 2




One piece rate: 1500 items per machine hour
Higher piece rate: 2000 items per machine hour
Higher yet: 2500 items per machine hour
Above plus higher rate: 3500 items per machine hour
No feedback, but incentives (new supervisor)
Reinstate feedback with Incentive 2 conditions
Last three phases represent a very nice
reversal, incentives with and without feedback
20
Union National Bank: Proof Operators
Fdbk
1,800
Baseline
1,065
~Industry standard
Inct 1
2,700
Inct 2
3,500
Incentive
w/o fdbk
Fdbk
restored
Why
reversal?
Notice:
Baseline in weeks
Remaining in months!
Almost 7 years of data!
(note that it took four months before performance declined w/o feedback - may not see it in shorter studies)
21
SO8: Effectiveness of feedback alone and with
tangible rewards, Balcazar et al.

% of articles in which performance consistently improved
 Feedback alone: ~30%
 Feedback with tangible rewards: ~90%

Take home point: Feedback works because it is
correlated with a differential reward system
But what about when it works alone?



Feedback may be linked to a differential reward system even
though that reward system was not explicitly designed by the
researcher/practitioner
Performance may have been measured for too short of time to
see a possible decrease over time - feedback would be expected
to increase performance temporarily when first introduced
22
SO9: Why feedback may be more
effective when provided by supervisor



SO9A: First reason
Feedback on how well workers are doing may
prompt supervisors to provide differential
consequences for good and bad performance
Why? The job of the supervisor is to get things
done through their workers; thus, when workers
perform well, it reflects well on the supervisor
and the organizational rewards that the
supervisor gets (good performance is
reinforcing/rewarding to the supervisor)
So…
23
SO9A, cont.
For the supervisor:
Antecedent:
Good feedback
R –––––––––––
Praises worker
> Sr
“Thanks!”
More good work
Antecedent:
Bad feedback
R –––––––––––
Criticizes worker
Prompts worker
> Sr
“I’ll do better”
Better work
Thus, feedback may change the consequences that the
supervisor provides to the workers.
(demonstration of this point: Richman et al. article, self vs supervisory feedback)
24
SO9B
SO9B: Second reason: Will be provided in lecture
25
SO10: Balcazar et al.’s fundamental conclusion

If no system of functional, differential consequences
exist, there is probably no point in establishing a
feedback system. Effort would be better spent
developing procedures (contingent rewards) for wanted
behaviors

In other words, you should not develop and implement a
feedback system until and unless you develop and implement a
system of functional, differential rewards! (or determine one
already exists)

This has become a problem in the field

37% of articles reviewed by Balcazar et al. used feedback alone
(review covered 1974-1984, approximately)

29% of articles reviewed by Alvero et al. used feedback alone
(review covered 1985-1998)
(This is straightforward, yet students seem to have trouble with it: more on next slide)
26
SO10: Implications of that fundamental
conclusion (NFE)

That also means that the effects of different feedback
characteristics may well depend upon whether or not
feedback is correlated with valued rewards and the
extent to which it is correlated with such rewards


Kang, Oah & Dickinson (2003) found that more frequent
feedback increased performance when individuals were paid
monetary incentives but not when they were paid hourly
This is a problem in organizations - it is often difficult to
determine the extent to which feedback is linked to
differential rewards in organizational settings, which
could account for discrepancies in the results of
feedback studies conducted in different organizations.
27
SO11: If feedback is established independently of
functional, differential rewards, what type appears to be
the most effective?

Feedback that is:

Graphic




Confirmed by Wilk & Redmon (1998)
Confirmed by Austin, Weatherly, & Gravina (2005)
Provided at least once a week
Combined with tangible rewards

This is interesting - basically the authors are saying you
need to implement a reward system along with the
feedback, so I am not at all sure it is a logical answer to the
question that was asked, but you should include it in your
answer.
(my interesting experience with graphic feedback in lab vs. written fdb).
28
SO12: Dickinson’s analysis of feedback


When most analyze feedback as a consequence, they
refer to it as a reward or positive reinforcement
contingency
Dickinson believes that feedback probably functions as
a negative reinforcement contingency, escape and/or
avoidance (which people refer to as an aversive
contingency)


Agrees with Malott and Michael on this one - “the world lives
by fear.”
Likely rule, “If I don’t perform well, I am going to be criticized
by my supervisor” (and/or am not likely to get what I want from
the supervisor)
(cont. on next slide)
29
SO12, cont: Why does Dickinson believe feedback
controls behavior via negative reinforcement?

As yourself these questions:





What types of consequences are most common in
business, industry, and human service settings?
That is, what typically happens when a person does
a good job?
What typically happens when a person does a bad
job?
Finally, why do organizations typically start to
measure performance in an organization?
Where do the answers to these questions lead?
Feedback controls behavior via escape and/or
avoidance contingencies
(something is not right -one more issue)
30
Posted individual and social comparison
feedback (NFE)
Over the years, some researchers have posted
individual data that is identified by code (see Anderson
et al. article this unit), individual data identified by
employee name (Anderson, Geller, Ludwig), and
individual data plus social comparison data (individual
and group).
These procedures have been very effective, with and
without supplemental consequences.
 What do you think about those type of procedures?
 What do you think is controlling performance when
they work?
31
Posted individual and social comparison
feedback: or Dickinson’s soap box (NFE)
Dickinson’s soap box will be provided in lecture
32
SO13: Richman et al. article, intro


Excellent example of PM research in a human
service setting
Human service settings are difficult from an
OBM perspective



More labor intensive due to required staff-client
contact to maintain and protect welfare of clients
Must focus more on behaviors of staff rather than
accomplishments - most of the work of the staff
consists of interactions with the clients
Staff behaviors and interactions can be naturally
punished by client behaviors (I’ve seen quite a few
human service workers with bruises and scratches!)
(I chose this article for two main reasons: the measurement system and the issues it raises about self-monitoring alone)
(I am just going to touch on a few of the objs over the articles: business dismisses work in human services-it’s
much more difficult to implement a PM program in a human service setting than in business)
33
Richman et al. intro, cont. (NFE)


Rationale of study
To determine whether a self-monitoring
procedure, with minimal supervisory
involvement, could increase staff adherence to
scheduled activities and on-task behavior
Participants
10 staff members in two houses of an
intermediate care facility for “mentally retarded”
(developmentally disabled)
34
SO13A: What two general categories of
behaviors were recorded?

On-schedule behavior




Is the staff member in the assigned area for the scheduled activity
according to the posted schedule?
Does the staff member have all of the materials necessary to conduct
the activity as indicated on the activity card?
13B Regardless of whether the staff member was actually
implementing the task (that is, the staff member could be off-task in
the sense of chatting with another staff member, drinking coffee, or
just interacting “generally” with the clients)
On-task behavior


Is the staff member engaged in behaviors for any of the three
appropriate activities (group, client/house custodial, or one-on-one
training)
13C Regardless of whether the staff member was implementing the
specific activity that had been scheduled (in other words, even if the
staff member was doing group training when one-on-one training was
scheduled)
(very nice measures of behavior, simple)
35
SO18: Self monitoring increased performance substantially.
Why, then, was supervisory feedback added?
Experimental Phases
Measures/
House
Baseline
In-service
Self-Monitor Self-Monitor
+ Feedback
A
50%
50%
80%
94%
B
39%
39%
75%
81%
A
28%
36%
72%
88%
B
28%
28%
77%
80%
On-schedule
On-task
(First, note lack of effectiveness of in-service - again; yes, they did get further increases, but the main reason -next slide)
36
SO18: Why add supervisory feedback?


The behavior of 5 of the 10 staff members became
variable over time (that’s 50% of the participants)
Supervisory feedback improved both on-schedule and
on-task behavior for each of the 5.


Demonstrates the importance of supervisory feedback
Also suggests that self-monitoring may be effective on
a short-term basis but may not be effective long-term



But, why would we expect self-monitoring to be effective
over the long run?
What consequences are there for self-monitoring or for the
self-monitored performance?
Mistake that many make in the field, just like with feedback
- that is assuming it will work without careful analysis of the
consequences/contingencies
(question: are we doing more harm than good when we publish short term studies that indicate that interventions are
Successful, particularly when the results don’t seem to conform to a solid behavior analysis?)
37
Anderson et al. article, one of Hantula’s first
studies

Fun study, because it was done in a student-run
university bar
“Cleaning was an ancillary requirement and also preempted time that could be spent with peers or studying.
The result was a conspicuous, pervasive accumulation
of grease and various sticky materials on virtually
every surface. Garbage areas were strewn with
debris.”
University officials said the bar had been a nearimpossible management situation for years, and the
State Board of Health threatened closure.
(again, I am just going to go over a few selected SOs)
38
Anderson et al. study, cont.




My main purpose for including it is that it is one of the
rare studies that has examined task clarification by
itself
Given the popularity of task clarification right now, it is
important that you understand that task clarification
alone may increase performance moderately, but you
should not expect large increases
That should not be surprising - task clarification is an
antecedent
However, it can be a very important part of a
component intervention - unless workers know what
they are supposed to do, they cannot do it!

In other words, I am not “dissing” task clarification!! (but I am
saying you probably should not use it alone)
(probably getting the idea by now from what I have said about feedback, self monitoring and now task clarification, I am a firm
believer in consequences of behavior.)
39
SOs 22 and 24: Effects of task clarification
and task clarification plus feedback


Task clarification increased task completion
13% over baseline
Task clarification plus feedback increased task
completion 62% over baseline
40
Crawley et al. article, introduction




I have included this article because it is the best one I have ever
seen with respect to improving sales behaviors and I would
wager that most many of the behaviors identified in their exquisite
analysis would generalize to other sales positions
Study was conducted by one of Ed Feeney’s consultants, Bill
Crawley (I didn’t stress Feeney’s accomplishments in U1, but I
recommend that you go back and read the Dickinson article for an
historical perspective)
The site was Ethan Allen
(NFE) Note the analysis at the beginning that was designed to determine
the best opportunities for intervention, based on both the potential for
improving performance (exemplar performer vs. average performer) and
the economic pay-off of intervening on the performance
41
SO25: What approach was not successful in
identifying what made sales reps effective?


Surveys were sent to the top sales representatives in the country
asking what is was that they did that made them so effective
This approach did not work because sales representatives could
not describe the behaviors that made them successful





I am friendly
It’s genetic - my parents were sales representatives
You need to be “up”
You need to be aggressive
General point
Even though workers are exemplary workers, they can’t tell you what
they do that makes them exemplary workers. Those behaviors are often
contingency-shaped (controlled by direct-acting contingencies) and they
never have had to describe them (describing what you do and doing
what you do are different behavioral repertoires)


Automobile mechanics
Construction workers
42
SO26: Exquisite specificity of the targeted
behaviors

To determine the behaviors




They observed 65 top performers
Over a four month period of time
For 1,000 hours
Both inside the store and at in-home sales calls




Recorded the stimulus-response sequences
That is, what were the antecedents that prompted a response by
the sales representative, and how did the sales representatives
respond to those antecedents
Also interviewed customers for 50 hours
(NFE) Pilot tested the entire intervention in two stores


First with the consultant as the coach
Then with the store manager as the coach (fidelity - did they
create an intervention that could be carried out by employees)
43
SO26: Exquisite specificity of the targeted
areas and behaviors, cont. (NFE)

Areas with 5-11 behaviors in each area








Customer greeted
Customer needs identified
Needs matched to store product and service
benefits
Objections identified and overcome
Decision maker identified
Close made
Results of sales contact
Follow-up action taken
(48 behaviors in addition to smiling, eye contact, natural voice, and use of customer’s name in each area)
44
SO26: Exquisite specificity of the targeted
areas and behaviors, cont. (NFE)

Examples of behaviors in the Customer Greeted area





Customers should be approached within 120 seconds after
entering the store
The sales representative should stand within 3-4 feet of the
customer, smile and maintain eye contact
The sales representative should approach the customer at a
normal pace and maintain a natural and relaxed posture
Introduce self using first and last names and identify his/her
position
Obtain the customer’s name and use it throughout the sales
interaction
45
SO28: Two reasons why commissions did
not function as effective rewards
Sales representatives received sales commissions
monthly which most would assume would be
sufficient to maintain high levels of performance

Commission payments were delayed, often by as many as 3
months, weakening the relation between sales and the
amount of money earned


Commissions earned in January would not be received until
March or April
Commissions were based on sales, an accomplishment
measure, and sales representatives did not know the
behaviors required to improve sales

The initial survey that failed to identify the critical target
behaviors showed that sales representatives did not know what
behaviors led to improved sales
46
SO29: Why is it important to compare data to records for
the same months in the preceding year?

As part of the analysis to determine the effectiveness of the
program, they compared the sales data to sales records for the
same month the preceding year. Why?
 Sales fluctuate seasonally and monthly
 February is traditionally a big sales month while December is
traditionally a low sales month
 In behavior analysis, we often use time series data (AB
design) to determine the effectiveness of our interventions but
 If you compared sales in February and it had increased in
comparison to Dec and Jan, you may conclude that your
program was successful when it was not
 Alternatively, if you compared Dec data with Nov data
(with traditionally higher sales), you may conclude your
program was not successful, when indeed it was
(including this just so you don’t just say “due to seasonal fluctuations” but add an explanation)
47
Questions over study objectives?

In-class exercise

Feedback can never be considered a reinforcer
because it does not always produce an increase in
performance



Agree or disagree?
Why?
In Komaki’s study feedback was not a reinforcer
because performance improved immediately upon
the introduction of feedback, rather than gradually


Agree or disagree?
Why?
48
Download