anchoring and consistency (keith dowd)
TRANSCRIPT
Anchoring and Consistency: A New Metacognitive Approach to Anchoring and Consistency: A New Metacognitive Approach to Understanding the Anchoring-and-Adjustment EffectUnderstanding the Anchoring-and-Adjustment EffectKeith W. Dowd John V. PetrocelliKeith W. Dowd John V. Petrocelli Wake Forest University Wake Forest UniversityWake Forest University Wake Forest University
ABSTRACTABSTRACT
The purpose of the present research is to test the hypothesis that
a metacognitive consistency factor is a mechanism that drives the
anchoring-and-adjustment heuristic. In a preliminary study, we
asked participants to respond to a series of trivia questions
requiring them to generate estimates based on a) their prior
knowledge (self-generated anchor; SGA) and b) a value that
functions as a marker to adjust upward or downwards from
(experimenter-provided anchor; EPA). We predicted that one’s
final estimate would be adjusted away from the EPA when the
consistency between the SGA and the EPA is low. In addition,
we hypothesized that SGAs play an important role when
estimating unknown quantities and that participants would be
aware of their SGAs, regardless of whether they were
consciously instructed to report them. Support was found for both
of our hypotheses which suggests that consistency may play an
important role in the cognitive processes underlying the
anchoring-and-adjustment heuristic.
INTRODUCTION
Anchoring-and-adjustment refers to a cognitive heuristic used in the
estimation process whereby people attempt to adjust their estimation
from an accessible starting value until an acceptable estimate is
reached; often the adjustment is insufficient (Kahneman & Tversky,
1974).
Previous research (who? Cite them) has demonstrated it to be robust
in and outside of the laboratory and has implicated it as one
explanation for numerical judgments and estimates of future
performance. Despite numerous studies chronicling its effect, little
research has explored the nature of the cognitive processes
underlying its operation.
Theoretical and Conceptual Framework
Epley & Gilovich (2007) suggested that some types of anchors (i.e.,
SGA) activate processes of anchoring-and-adjustment whereas others
do not (EPA). They argued that semantic priming is occurring in the
case of EPAs rather than anchoring-and-adjustment. We propose that
adjustment does occur with EPAs but that it does so in conjunction
with SGAs and is driven by a metacognitive consistency factor
between the two types of anchors.
Because people also have the ability to think more deeply about their
initial thoughts, we suggest that a metacognitive view (Nelson &
Narens, 1994) of anchoring-and-adjustment may be fruitful to our
understanding of its operation.
Goal• To test the hypothesis that consistency between SGAs and EPAs
moderates the degree of adjustment represented in final estimates
from SGAs after being exposed to the EPAs.
EXPERIMENT
HYPOTHESES
1. We predicted that final estimates would be associated greater
adjustment away from the EPA when the consistency between the SGA
and the EPA is low.
2. We also predicated that SGAs operate to influence final estimates
regardless of whether SGAs are consciously reported or not.
PROCEDURE
• Participants (N=84) completed 12 item sets• The question used in each item set was constructed such that
participants would not know the correct answer• Randomly assigned to 1 of 6 conditions that varied on:
• The experimenter-provided anchors provided (high v. low)• The request to self-generate (presence v. absence)• The location of the request to self-generate (first v. last)
• Each item set consisted of 4 sequential parts:
1. Presentation of the question and self-generation of internal anchor
(SGA)
2. Presentation of the experimenter-provided anchor (EPA)
3. Presentation of the comparative assessment question (CA)
4. Request for final estimate (FE)• 6 of the items served as filler questions• Dependent variable: Amount of adjustment in FE from EPA
SGASGA FEFEEPAEPA CACA
Conditions 1 & 2
SGASGA FEFEEPAEPA CACA
Conditions 3 & 4
SGASGAFEFEEPAEPA CACA
Conditions 5 & 6
RESULTS
• Average correlation between consistency and adjustment = -.xx (all ps <.0?)• Average correct detection of high/low EPAs = xx%• All 6 of our items evidenced either a significant main effect for consistency or
an interaction between the experimenter-provided anchors and consistency,
even after controlling for participants’ self-generated anchors as a covariate.
• Evidence suggests that SGA are generated spontaneously (no difference in
total adjustment between conditions 3/4 and conditions 1/2 and 5/6).
• The temporal location in the items sets of the request for participants to form
self-generated anchors did not have a significant effect on their final estimates.
DISCUSSION
The results of our experiment suggest two ways in which a metacognitive
process may operate when people respond to classic anchoring questions.
One is the pattern of correct CA judgments made by participants; that is,
participants were largely accurate in determining whether or not the EPA
was high or low. This finding suggests that they are not only using a self-
generated standard of comparison (e.g., SGA) but also actively thinking
about it when making estimates about unknown qualities.
The other way in which our data support a metacognitive process comes
from the recurrent interaction that occurred between consistency and EPAs.
Participants appear to be consciously reflecting on the consistency between
their SGAs and the EPAs in order to a render a final estimate.
Our data also speaks to the spontaneity of SGAs. That is, these estimates
were generated by participants regardless of whether or not they were
instructed to explicitly report them; which suggests that they play an
important role in the operation of the anchoring-and-adjustment heuristic
because they provide a standard of comparison to contrast against EPAs.
The consistency between these two anchors moderates the amount of
adjustment in FEs and without the contribution of both of them the anchoring
effect may fail to operate.
Finally, these data build on the findings of Epley & Gilovitch (2007) because
they provide further evidence for the importance of SGAs for the anchoring
heuristic while also offering a preliminary process account for its operation.
Figure 1
Adjustment from experiment-provided anchor means by experiment-provided anchors and
consistency level between self-generated and experiment-provided anchors.
U N I V E R S I T Y
WAKE FOREST