Cognitive psychology is a branch of psychology that examines the mental processes of perception, thinking, memory, attention, decision-making, problem-solving, language use etc. Decision-making can be defined as ‘the cognitive process of choosing between two or more alternatives’ and thinking is ‘a cognitive behavior in which ideas, images, mental representations or other hypothetical elements thought are experienced’ (APA, 2020). There are different approaches to thinking and decision-making. The model of normative decision theory assumes that the decision-maker acts on the basis of a series of logical implications and thus makes the most rational decision. However, the model of descriptive decision theory explains the decisions that people who are influenced by emotions and deviate from norms are prepared to make. The dual-process theory of reasoning (Kahneman, 2003) has been studied many times with different approaches. The general assumption is that thinking is an interaction between two systems. The heuristic system, which is based on prior knowledge and beliefs, and the analytical system, which is based on current logical standards. The analytical system relies heavily on computations and is successfully engaged when enough resources are available. The two systems can sometimes cue different responses (Wim De Nays, 2006). Heuristics are the simplified mental strategies that people use to reach a decision or solve a problem when there is no time or resources available for further analysis. The use of heuristics can lead to cognitive biases, systematic deviations from rationality in judgment, or decision making (Blanco, 2017). The anchoring bias is a heuristic identified by Kahneman and Tversky. The anchoring heuristic consists of the assessment of a probable value of an event or result, with the assumption being based on a kind of a reference point (an anchor). It often leads to an inappropriate evaluation resulting from the inability to let go of the anchor (Zimbardo & Gerrig, 2012). Investigating the anchoring effect helps to understand how System 1 is utilized in thinking and decision-making.
Strack and Mussweiler conducted a series of three studies in 1997 which aimed to investigate whether pre-generated information (anchors) influenced decision-making and were therefore a determining factor in anchoring bias. About 60 German non-psychology students were recruited to participate in the study. The participants received a questionnaire on "general knowledge", which consisted of a pair of questions. They were first asked for a comparative and then an absolute judgment. For example, they were first asked whether Gandhi was younger or older than 9 (or 140) years old at the time of his death. Then they were asked about the specific age at which Gandhi died. The participants in the low anchor condition (9 years old) assumed on average that Gandhi was 50 years old when he died. In contrast, in the high anchor condition (140 years old) 67 was the average answer. The results of the study prove the anchoring bias - the numerical information given to the participants in advance influenced the absolute judgments about Gandhi's age at the time of his death. Both of the anchors were implausible, but the low anchor had a greater impact on the results of the study.
Similarly to the study carried out by Strack and Mussweiler in 1997, an experiment investigating the anchoring bias and its influence on decision-making was conducted. Anchors can have an effect on human behavior in everyday life - in decision-making and problem-solving e.g. in sales. The influence of the anchoring bias was tested with a general knowledge quiz, consisting of five questions one of which was vital to the study, while the rest were filler questions (inspired by the original study). The independent variable was the low or high anchor in the crucial comparative question: ‘Was Gandhi older or younger than 9 (low anchor)/140 (high anchor) when he died?’. The dependent variable was the estimated age in the absolute judgment question: ‘How old was he exactly?’ and it was measured in numerical values given by the participants.
Aim - to investigate whether pre-generated information (anchors) will influence the absolute judgment
Experimental hypothesis - The anchors will influence the judgment about Gandhi’s age at the time of the death, by either lowering (in the low anchor group) or raising (in the high anchor group) the estimate.
Null hypothesis - There will be no difference in the estimates of Gandhi’s age at the time of his death in relation to the low or high anchor in both groups or any difference will be due to chance.
An independent measures design was used in order to avoid demand characteristics on the participants’ part. The independent variable was manipulated by randomly allocating participants into groups. All potential confounding variables canceled each other out.
Opportunity sampling was used due to high accessibility. The sample consisted of 22 high school students of more than 1 gender, aged 16-18. The participants were randomly allocated into either the low anchor condition or the high anchor condition. The conveniently obtained sample could be not representative of the total population of Polish high school IB students. The sample was chosen due to convenience and the interest in the influence of anchoring bias on the researchers’ peers.
Every person received detailed instructions. All participants were required to sign an informed consent form. Every individual participant received a 5-question quiz, the quizzes were distributed randomly. The participants were asked to fill out the questionnaire to the best of their abilities or provide their best guess. The time for answering the questions was unlimited. After the experiment, the standardized debriefing was applied via email.
All the collected results were collected in one table (see Appendix E). The estimates in the low anchor group ranged from 11 to 86 (R=75). The range in the high anchor group started at 40 and ended at 101 (R=61). The results are presented in the form of the mean value, as the collected data is representative of ratio data - numerical values with a true zero. The processed results themselves are very remote and suggest the acceptance of the alternative hypothesis.
Low anchor | High anchor |
---|---|
47 | 78 |
Figure 1 - Table On Mean Value Of The Estimates
Low anchor | High anchor |
---|---|
26 | 19 |
Figure 2 - Table On Standard Deviation Of The Estimates
The hypothesis proposed was directional and one-tailed and ratio data was obtained. An independent measures design was used. Therefore, an unrelated t-test was used (a parametric statistical test)
Degrees of freedom = 20. The t-value is -3.52007. The p-value is .001017. The result is significant at p < .05.
The critical value for a one-tailed hypothesis with 20 degrees of freedom at p-value 0.05 is 1.725. Therefore, we can accept the experimental hypothesis, reject the null hypothesis and conclude the result as statistically significant.
The results showed how either a low or a high anchor influences the final judgment. There was a significant difference in the means between the two conditions. (low anchor = 47; high anchor = 78) the mean results were much more pronounced than in the original study by Strack & Mussweiler from 1997 (low anchor = 50; high anchor = 67). The low anchor lowered the final estimate, while the high anchor raised the mean estimate. The unrelated t-test proved the significance of the results at p <.05. In the original study, the significance of the results was proved at p < .02. An unrelated t-test was used to test the significance of the results. The data was representative of ratio data and the hypothesis was directional one-tailed (estimates are lower in low anchor conditions but higher in high anchor conditions). Under these conditions, the Mann-Whitney U test could also be used, but the data obtained were normally distributed and not continuous. Therefore, an unrelated t-test was chosen, despite the lack of similarity in the variances between the groups.
The use of the independent measures design was necessary to investigate both of the experimental conditions (low anchor and high anchor groups). The use of the independent measures design was very time efficient, as practically all participants were able to complete the questionnaire at the same time. What we could have done is to include a control condition without an anchor to have a reference point for the low anchor group and the high anchor group. The participants could not be aware of the anchor question used in the other group, so of course order effects - distortion of the results by boredom, fatigue, practice caused by presenting the independent variables in a certain order (APA, 2020) - were eliminated. One limitation of the design was that only the influence of one anchor was applied to each participant. If a repeated measures design were to be used, the influence of both anchors could be investigated in each participant.
Opportunity camping was chosen for convenience. The sample was obtained quickly and easily. However, all participants attended the same high school. This variable could falsify the results, making the sample not representative of the whole target population. Besides, the educational level of the students could be similar, so that their intellectual level could change the results. If the experiment were to be repeated hypothetically, it would be profitable to use a random sample.
Overall, the experiment could be conducted with a more diverse sample. More pressure could be applied to control the likely confounding variables, e.g. noise. The filler questions could be more similar to the anchor question since the anchor question was the only one that contained two parts. A desirable modification would also be the introduction of complete silence (some participants could be distracted, talk to themselves, etc. even if they are asked to remain silent). More anchor questions could be included in the quiz to exclude the possibility of demand characteristics - participants interpreting the aim of the research (the anchor questions could be different from the others).
The experiment carried out is a good illustration of the mechanism of the anchoring effect. They explain how system 1 activates intuitive, automatic thinking, which results in the anchoring effect. The results are consistent with the theory and the original experiment by Strack and Mussweiler from 1997. It is clear that the high anchor caused participants to increase their assumptions and the low anchor caused participants to decrease their estimates. These results support the anchoring heuristic hypothesis.
I confirm that -
Therefore, I give my consent to participating in the study
Thank you for your will to participate in the experiment. Read the consent form carefully and sign adequately to your decision. If you happen to change your mind, you should signal one of the researchers immediately. Remain as focused as possible throughout the whole procedure. Your task will be to answer 5 general knowledge questions. Please put the answers on a given sheet of paper. If you have any questions, ask the researchers.
When was Leonardo Da Vinci born?
What is the mean temperature in the Antarctic in the winter?
How long is the Mississippi River?
Was Gandhi older or younger than 9/140 when he died? How old was he exactly?
What is the width of the Brandenburg Gate?
Thank you for participating in our study. The research was an investigation of a phenomenon, which causes people to rely heavily on a piece of information given beforehand, called the anchoring bias. Within the questions you answered, the one about Gandhi's death consisted of either a low or a high anchor - to see if it would influence your answer. Let us know if you want to find out about the results of the study!
Low anchor | High anchor |
---|---|
68 | 101 |
86 | 96 |
64 | 94 |
70 | 40 |
20 | 70 |
56 | 75 |
15 | 62 |
19 | 81 |
11 | 97 |
46 | 80 |
58 | 62 |
Figure 4 - Table On Raw Data