The Decision Sciences Journal of Innovative Education

To guess is inexpensive; to guess wrong can be very costly.

                                                       Ancient Chinese proverb

In this teaching brief, we describe a technique  for demonstrating how cognitive heuristics subtly (and sometimes perniciously) affect decision-making.  More specifically, we describe a method for illustrating the anchoring effect.  Awareness of this effect is, logically, the first step toward obviating and/or removing the potential biases that may result from basing decisions on unreliable/irrelevant information. [1]

Cognitive heuristics are mental mechanisms used to cope with the uncertainty and complexity of the decision-making environment (Bazerman, 2002).  By reducing the amount of information taken into consideration, heuristics simplify and speed up the decision-making process (Schwenk, 1986).  But the heuristics employed can have negative consequences, resulting in sub-optimal decisions. 

Anchoring is one such heuristic, where managers make a decision based on an initial parameter or parameters (Tversky & Kahneman, 1974).  The initial value(s) may be drawn from memory or experience, or may be supplied by others. When based on irrelevant or no longer pertinent information, faulty decisions are likely to result.  When making forecasts, people often use the past as the starting point.  While the past may be relevant, the environment may offer other pertinent clues to the future.  Illuminating potential anchoring biases may enable decision-makers to consider what information they are considering.

            Bazerman (2002) discussed a model for improving decision-making that builds on Lewin’s (1947) framework of unfreezing, change, and refreezing.  Making people aware of a potential anchoring bias may allow them to unfreeze their decision-making process.  If anchoring is successfully demonstrated, the decision maker is likely—at least in future decisions—to consider the possibility that s/he has permitted a potentially irrelevant or incorrect piece of information to anchor or bias the decision.  This revelation may thereby result in a change in his/her decision-making behavior.  If that change proves to be successful, refreezing will likely occur as the decision-maker incorporates knowledge of anchoring into future decisions. 

The teaching of cognitive heuristics is particularly relevant to courses in organizational behavior and managerial decision-making.  Not only can this knowledge be useful to future managers in improving their own decision-making abilities, but it can also illuminate why bosses and subordinates make the decisions they make.  But simply telling (i.e., lecturing) students about cognitive heuristics is not likely to be enough.  An active demonstration of the anchoring bias that personally involves the students is more likely to unfreeze the decision-making process.

Bonwell and Eison (1991, p.2) define active learning as “anything that involves students in doing things and thinking about the things they are doing.”  Sivan, Leung, Woon, and Kember (2000) demonstrated that active learning promotes the development of critical thinking and problem solving, two important skills for decision-makers at all levels.

Oslund, Rubin, and Kolb (2000), in the context of higher organizational behavior, and Silberman (1998) in connection with training, recommend combining abstract, conceptual learning with active problem solving in the teaching of organizational psychology. Making concepts pertinent to students’ future employment by giving concrete illustrations enhances the relevance of the course content.  Our demonstration incorporates an active learning approach along with a conceptual explanation of the anchoring effect.

THE EXERCISE

The exercise is conducted in the form of a problem set completed in class.  All participants are given a scenario in which they are asked to make a judgment. The scenario includes information that is intended to be irrelevant to the judgment at hand, but which may serve as an anchor nonetheless.  On a random basis, one-half of the students receive the low anchor scenario; the other one-half receive the high anchor scenario.  Students are not informed about the existence of two scenarios.  Upon completing the problem set, students are directed to transfer their answers to the last page (the answer page), and to tear it off and pass it along to the teaching assistant (or to volunteer student scorers) for scoring.  The two versions of the problem appear below:

Here is the low anchor version, labeled Problem 2:

A newly hired engineer for a medical laser firm in San Diego had four years of experience in excimer lasers and excellent qualifications.  When asked to estimate the starting salary for this employee, my administrative assistant (knowing very little about the profession or the industry) guessed an annual salary—excluding benefits—of  $45,000.  What is your estimate?

Here is the high anchor version, labeled Problem Two:

A newly hired engineer for a medical laser firm in San Diego had four years of experience in excimer lasers and excellent qualifications.  When asked to estimate the starting salary for this employee, my administrative assistant (knowing very little about the profession or the industry) guessed an annual salary—excluding benefits—of  $145,000.  What is your estimate?

            The job of excimer laser engineer was selected because it was reasoned that very few students would have accurate information as to salary.  The location of San Diego was selected so that all participants used the same location as a basis for their estimate.  The executive assistant’s opinion is the only different piece of information received between the two groups of students.  Note that both versions stipulate that the executive assistant does not have any knowledge on the salary of the excimer laser engineer.  (Results would likely have been more powerful without this disclaimer.)

During the 15-minute period while the professor gives a brief lecture on various forms of cognitive heuristics and their effects on decision-making, the mean scores are calculated for the two groups.  (The answer page allows the identification of the scenario condition: the label “Problem 2” is used in the low anchor group; the label “Problem Two” denotes the high anchor condition.)   Typically, the lecture includes the biases emanating from the representativeness heuristic (e.g., regression to the mean, the conjunction fallacy), the availability heuristic (e.g., ease of recall, vividness), and the anchoring/adjustment heuristic (e.g., overconfidence, anchoring effect) (Bazerman, 2002).  When it comes time to debrief the exercise (but prior to telling them about the fact that there are two different versions), it is instructive to ask a small group of students to indicate their answers, and how they arrived at them. When asked, students will say that they did not take the secretary’s estimate into account when forming their own judgment. Next, when discussing the potential bias of cognitive anchors, the instructor reports the means in the two conditions.  Students subsequently are surprised by the existence of two conditions, and look at each other’s problem sheets to verify the existence of differing anchors.

RESULTS

            The above technique has been used in thirteen separate trials over a period of 10 years.  Participants have been both graduate and undergraduate students attending public and private institutions. In all trials, the mean for those given the high anchor point has exceeded the mean for those given the low anchor point.  Mean scores across the 13 administrations (unweighted) were $99,790 and $68,860 in the high and low anchor conditions, respectively, t = 7.45 (p < .001).  Put differently, the average estimate in the high anchor condition is approximately 45% greater than the average estimate in the low anchor condition!  Remarkably, this demonstration has never failed to elicit a sizable anchor effect, which is all the more compelling in light of the fact that many students informally say that they discounted the anchor (i.e., the secretary’s estimate) when forming their own judgment.  Students are surprised and impressed by the magnitude of the effect demonstrated by this exercise and it piques their interest in the effects of other sources of decision bias.  Presentation of the results to date seemingly eliminates skepticism that the present results are an anomaly.

DISCUSSION

            Clearly there has been a reliable short-term effect.  A long-term evaluation assessing the effectiveness of this demonstration should also be performed, possibly comparing students who have participated in the anchoring demonstration to those who have merely read or heard about the anchoring effect. 

            Some educators are reluctant to use active learning exercises in general: classroom time is limited; outcomes are uncertain.  This exercise can be counted on to succeed—at least until “word of mouth” communication among students diminishes the pool of students unaware of the anchoring effect.

Finally, it is worth noting that the present problem set protocol could be expanded to demonstrate other decision biases.  For example, a scenario might be developed where a building initially purchased for $400,000 (by one-half the class), or $100,000 (by the other half), is now worth $200,000.  The fact set could permit a tax-free swap into a superior building worth $200,000 or the present building might be held until it is worth $400,000, and swapped for an even better building.  This situation would illustrate how the consideration of sunk costs may confound the decision process.

REFERENCES

Bazerman, M.H.  (2002).  Judgment in Managerial Decision Making, 5^th  ed.  New York:

     John Wiley & Sons.

Bonwell, C.C. & Eison, J.A.  (1991).  Active learning:  Creating excitement in the classroom. 

     ASHE-ERIC Higher Education Report No. 1.  The George Washington University, School

     of Education and Human Development, Washington, DC.

Lewin, K.  (1947).  Frontiers in group dynamics: Concept, method, and reality in social science;

     social equilibria and social change.  Human Relations, 1, 5-41.

Oslund, J., Rubin, I.M., & Kolb, D. A.  (2000).  Organizational behavior: An experiential approach.  Englewood Cliffs, NJ:  Prentice-Hall.

Roxburgh, C.  (2003).  Hidden flaws in strategy.  McKinsey Quarterly, Spring 2003(2), 26-40.

Schwenk, C.R. (1986).  Information, cognitive biases, and commitment to a course of action. 

     Academy of Management Review, 11, 298-310.

Silberman, M. (1998). Active Learning, 2^nd Ed. San Francisco: Jossey-Bass/Pfeifer.

Sivan, A., Leung, R.W., Woon, C., & Kember, K.  (2000).  An implementation of active learning

     and its effect on the quality of student learning.  Innovations in Education and Training

     International, 37(4), 381-389. 

Tversky, A. & Kahneman, D.  (1974).  Judgment under uncertainty:  Heuristics and Biases.

     Science, 185, 1124-1131.

_____________________

[1] The authors thank Professor Joel Brockner, Columbia Business School, and two anonymous reviewers for their helpful comments.