Marginal Dishonesty: The Adding-to-10 Task
6,209 Views
•
•
Vue d'ensemble
Source: Julian Wills & Jay Van Bavel—New York University
Classical economic theory asserts that people are rational and self-interested. In addition to seeking wealth and status, people are motivated by other goals. As a result, financial motives can sometimes be dwarfed by other internal needs, such as maintaining a positive self-concept or affiliating with other group members.
Ethical dilemmas, such as the temptation to cheat on taxes, can result when these motives are in conflict. On the one hand, people may be tempted to save money by underreporting their taxable income. On the other hand, no one wants to perceive themselves as a dishonest, free-rider. As a result, people are reluctant to fully exploit unethical opportunities because doing so can severely undermine their self-image as morally upstanding individuals. Instead, people cheat to a much smaller degree than they are capable of: just enough to gain additional resources, but not so much as to compromise their self-image.
This tendency for marginal dishonesty, or the “fudge factor,” is an important principle in social psychology and can be tested through a variety of techniques. Mazar, Amir, and Ariely originally described six separate experiments involving (dis)honesty and a theory of self-concept maintenance.1 The “Adding-to-10 Task” is one of the experimental techniques discussed and is prevalent in research that involves testing honesty. This video demonstrates how to produce and interpret the Adding-to-10 Task.
Principles
Principles of honesty are rooted in the philosophies of Thomas Hobbes and Adam Smith. Modern economic models espouse the belief that people behave dishonestly by consciously weighing the benefits versus the costs of the dishonest acts. This cost-benefit analysis considers possible external rewards, the probability of being caught and the magnitude of possible punishment. Psychologists build upon the economic model by introducing the effect of internal rewards. When people comply with their internal values systems, derived from society norms, they are provided with positive rewards, whereas noncompliance results in negative rewards, i.e., punishment. This internal reward system affects people’s self-concept, their self-perception which is influenced greatly by notions of morality.
Procédure
1. Participant Recruitment
- Conduct a power analysis and recruit a sufficient number of participants.
- Randomly assign half the participants to the experimental condition and the other half to the control condition.
2. Data Collection
- Give participants a test booklet with twenty matrices from the Adding-to-10 Task.
- Each matrix is based on a set of twelve three-digit numbers, two of which sum exactly to 10 (see Figure 1 for example).
- This task is beneficial because the answers are unambiguous.
Figure 1: One of the more common test stimuli used to elicit the fudge factor is the Adding-To-10 Task. Participants are instructed to find two numbers that add to ten in each matrix (e.g., 4.31 and 5.69 in the example above).
- Inform participants that, at the end of the session, two randomly selected participants will receive a bonus payment of $10 for each correctly solved matrix.
- Explain to participants that their goal is to circle the two numbers on each matrix that add to 10 and to complete as many as possible within 4 min.
- It is imperative that the test is challenging enough so that most participants are unable to correctly answer all questions in the allotted time.
- Call time after 4 min and instruct the participants to stop writing.
- For the control condition: Collect test booklets directly from participants. Verify and record the number of questions correctly answered.
- This will ensure that participants in the control condition have no opportunity to cheat.
- In the experimental condition: Read the correct answers to participants and allow them to 'grade' their own performance.
- Instruct them to tear off the back blank page of the booklet and write their name and number of total correct answers.
- Further instruct them to leave their answer page on the front desk and then dispose of, or take with them, the booklet.
- This provides the experimental group with an opportunity to cheat since the answers they actually recorded in the booklets cannot be verified.
- Fully debrief participants.
3. Data Analysis
- For the dependent measure, calculate the performance of both conditions by counting the number of correctly answered questions (control condition) versus the number of correctly answered questions reported(experimental condition).
- The control condition provides a baseline estimate since there is no opportunity to cheat. If people exploit the opportunity to cheat, then the number of correct answers reported in the experimental condition will be larger in comparison.
People inherently want to reap the benefits of cheating, even if they view themselves as honest.
For example, a salesperson making a commission on every car they sell might be tempted to cheat, reporting that they sold more vehicles than they actually did. On one hand, they’ll consider the costs of this action: whether they’ll be caught and punished by an employer.
However, external rewards—like how much extra money could be made—and internal rewards—whether they can still view themselves as an honest person—also influence this decision.
As a result of this interplay between costs and rewards, many individuals might opt to be marginally dishonest—noting that they sold only a few more vehicles, rather than the entire lot. This way, they still get the benefit of having some extra cash, but their moral self-image isn’t too adversely affected.
This video explores the relationship of just how much people cheat when a cash prize is rewarded as demonstrated by the Adding-to-10 task, in which two numbers that sum to 10 are identified from a given set.
In this experiment investigating honesty, participants are asked to complete math problems—the Adding-to-10 task—and depending on their group assignment, their answers are either graded by the researcher, in the case of the controls, or themselves—the experimental condition.
The math puzzles consist of 3 x 4 matrices containing numbers below 10. The trick is that only two of the values in a grid add to 10, and—to answer the problem—each component of this adds-to-10 pair must be circled.
Participants are given a booklet of 50 such matrices to solve as many as they can within 4 min, and told that random winners will receive $10 for every correct answer. Importantly, this potential cash prize is an external reward, which encourages dishonesty at a later stage.
For the control group, workbooks are collected and graded by the researcher immediately after time is up; there is no opportunity to inflate one’s results. In contrast, participants in the experimental condition correct their own work by listening to answers the researcher reads aloud.
Afterwards, they’re told to write down how many matrices they solved on a single piece of paper. It’s emphasized that only this sheet will be collected; the rest of the workbook can be thrown away.
The idea is that, since the accuracy of self-reported grades won’t be checked, there’s no cost to cheating. Thus, participants will be tempted to lie and claim that they solved more matrices than they actually did, all in the hopes that they’ll win a larger cash prize.
Here, the dependent variable for the experimental group is the number of matrices participants report they correctly answered. This can be compared to the number of matrices that were actually solved by control participants, which serves as a baseline value.
Based on previous research, it is expected that most experimental participants will inflate their results slightly, indicating they answered several more matrices than control individuals.
Such marginal dishonesty suggests that the internal reward of still being able to view oneself as a moral person prevents most people from lying excessively, and claiming that they solved all problems.
First, to calculate how many participants are needed, perform a power analysis. To begin, greet each one when they arrive. Then, provide them with a booklet of matrices, and explain the task.
Emphasize that to correctly answer a matrix, the two values in it that add to 10 must be identified and clearly circled. Also stress that a cash prize—$10 for every correct answer—will be awarded to two random winners once all of the data are collected.
Ensure that the participant understands the task, and then allow them to solve as many math problems as possible in 4 min.
When the time is up, have the participant put their pen down. For those in the control group, immediately collect their workbook. Check each of the matrices, and note how many were answered correctly.
For members of the experimental group, explain that they will be correcting their own work as matrix answers are read aloud. Then, proceed to recite the solutions.
Emphasize that the rest of the workbook won’t be collected, and should be thrown away when they leave. Then, tell the participant to write down their name and both the number of math problems they answered correctly and incorrectly. Afterwards, have them turn in this summary answer sheet.
Once the data have been collected, debrief each individual, and explain that this task is meant to investigate how the costs and benefits of cheating influence a person’s decision to be dishonest.
To analyze the data, calculate the number of correct answers and plot as a frequency distribution showing the percentage of participants by group.
Notice that individuals allowed to self-report their results indicated that they solved a significantly higher number of matrices compared to controls.
However, based on the overlap of the distributions, most participants in the experimental condition cheated just a little with marginal inflation, and only a small percentage cheated a lot.
This suggests that participants weigh two things in their decision to be dishonest: The first is the external reward of $10 per correct matrix, which results in their lying about the number of problems they solved.
The second is the internal reward of their moral self-image—still being able to view oneself as an honest person—that effectively caps this lie, and deters participants from claiming they solved the maximum number of problems.
Now that you’ve learned how the Adding-to-10 task can be used to explore the relationship between the extent of cheating and self-image, let’s take a look at other ways researchers are investigating dishonesty.
On one hand, some scientists are evaluating how moral priming—reminding an individual about honest concepts and actions—influences their willingness to cheat. Students were asked to sign an honor code contract, and afterwards were administered the Adding-to-10 task.
It was found that there was no significant difference between the number of matrices these experimental group participants noted they solved and data from control individuals.
This suggests that prompting individuals with their institution’s honor code reminds them to be moral, and decreases their tendency to cheat.
In addition, such reminders of honesty are being extended in other settings, like an office, where supplies often go missing. In this case, it’s been shown that if a mirror is placed in an area where individuals might be tempted to steal—such as in the supply closet—they are discouraged from taking home all the pens.
The idea is that, by observing their actions, the potential thief is made self-aware, and they realize that taking all of them might mean they’re not as honest as they’d like to believe.
You’ve just watched JoVE’s video on how the Adding-to-10 task can be used to better understand dishonesty. By now, you should know how to design the experiment, collect control and cheating data, interpret the results, and grasp how self-image—especially reminding an individual of their actions—could be used to discourage cheating in other applications.
Thanks for watching!
Résultats
This procedure typically results in a considerably higher number of correctly "solved" questions in the experimental condition (Figure 2). This procedure can also dissociate whether this inflated performance is a result of a few individuals cheating a lot or most individuals cheating a little bit. If the former were true, this would result in a mostly overlapping distribution except for a large relative increase of individuals reporting the highest possible score. Instead, typical results reveal that most participants cheat a little bit.
Figure 2: A typical frequency distribution resulting from the task. In this example, there is one experimental condition and one control condition with no opportunity to cheat. The y-axis values reflect the proportion of individuals who reported correctly solving a specific number of test questions. Values on the x-axis represent bins of three numbers centered on the label displayed (e.g., 30 = participants who solved 29, 30, or 31 questions).
Applications and Summary
People inherently are torn between achieving gains from cheating versus maintaining a positive self-concept of honesty. By using techniques like the Adding-to-10 Task, modern psychological research concludes that often people, who think highly of themselves in terms of honesty, will rationalize their behavior in such a way to allow them to engage in limited dishonesty while maintaining positive views of themselves. Put another way, there is an acceptable level of dishonesty that is defined by internal reward considerations. Given these factors, dishonesty may actually decrease as external rewards increase, i.e., the internal punishment does not kick in until a certain level of gain is achieved.
Economists estimate that dishonest behaviors (e.g., cheating on tax returns, returning clothing after use, employee theft, etc.) cost organizations billions of dollars each and every year. Legislative regulations that penalize dishonesty can be expensive and exploited. In contrast, research suggests that interventions that appeal to our motives for self-image maintenance may be cheaper and more effective. For instance, research suggests that subtly priming people's self-awareness (e.g., placing a mirror behind a jar of money) can reduce theft.2
These findings also cohere with one of the core tenets of social psychology: Almost everyone is capable of misbehaving depending on the situation. Efforts to discourage cheating might be more effective if they focus less on the rare master-mind criminal and instead address the possibility that most people cheat slightly. Interventions that draw attention to ordinary people's self-image may be fruitful for reducing this temptation. For instance, Mazar et al. found that priming participants with The Ten Commandments dramatically reduced cheating (even among atheists).
References
- Mazar, N., Amir, O., & Ariely, D. (2008). The dishonesty of honest people: A theory of self-concept maintenance. Journal of Marketing Research, 45, 633-644.
- Ariely, D. (2012). The (honest) truth about dishonesty: How we lie to everyone-especially ourselves. HarpersCollins. New York.
Transcription
People inherently want to reap the benefits of cheating, even if they view themselves as honest.
For example, a salesperson making a commission on every car they sell might be tempted to cheat, reporting that they sold more vehicles than they actually did. On one hand, they’ll consider the costs of this action: whether they’ll be caught and punished by an employer.
However, external rewards—like how much extra money could be made—and internal rewards—whether they can still view themselves as an honest person—also influence this decision.
As a result of this interplay between costs and rewards, many individuals might opt to be marginally dishonest—noting that they sold only a few more vehicles, rather than the entire lot. This way, they still get the benefit of having some extra cash, but their moral self-image isn’t too adversely affected.
This video explores the relationship of just how much people cheat when a cash prize is rewarded as demonstrated by the Adding-to-10 task, in which two numbers that sum to 10 are identified from a given set.
In this experiment investigating honesty, participants are asked to complete math problems—the Adding-to-10 task—and depending on their group assignment, their answers are either graded by the researcher, in the case of the controls, or themselves—the experimental condition.
The math puzzles consist of 3 x 4 matrices containing numbers below 10. The trick is that only two of the values in a grid add to 10, and—to answer the problem—each component of this adds-to-10 pair must be circled.
Participants are given a booklet of 50 such matrices to solve as many as they can within 4 min, and told that random winners will receive $10 for every correct answer. Importantly, this potential cash prize is an external reward, which encourages dishonesty at a later stage.
For the control group, workbooks are collected and graded by the researcher immediately after time is up; there is no opportunity to inflate one’s results. In contrast, participants in the experimental condition correct their own work by listening to answers the researcher reads aloud.
Afterwards, they’re told to write down how many matrices they solved on a single piece of paper. It’s emphasized that only this sheet will be collected; the rest of the workbook can be thrown away.
The idea is that, since the accuracy of self-reported grades won’t be checked, there’s no cost to cheating. Thus, participants will be tempted to lie and claim that they solved more matrices than they actually did, all in the hopes that they’ll win a larger cash prize.
Here, the dependent variable for the experimental group is the number of matrices participants report they correctly answered. This can be compared to the number of matrices that were actually solved by control participants, which serves as a baseline value.
Based on previous research, it is expected that most experimental participants will inflate their results slightly, indicating they answered several more matrices than control individuals.
Such marginal dishonesty suggests that the internal reward of still being able to view oneself as a moral person prevents most people from lying excessively, and claiming that they solved all problems.
First, to calculate how many participants are needed, perform a power analysis. To begin, greet each one when they arrive. Then, provide them with a booklet of matrices, and explain the task.
Emphasize that to correctly answer a matrix, the two values in it that add to 10 must be identified and clearly circled. Also stress that a cash prize—$10 for every correct answer—will be awarded to two random winners once all of the data are collected.
Ensure that the participant understands the task, and then allow them to solve as many math problems as possible in 4 min.
When the time is up, have the participant put their pen down. For those in the control group, immediately collect their workbook. Check each of the matrices, and note how many were answered correctly.
For members of the experimental group, explain that they will be correcting their own work as matrix answers are read aloud. Then, proceed to recite the solutions.
Emphasize that the rest of the workbook won’t be collected, and should be thrown away when they leave. Then, tell the participant to write down their name and both the number of math problems they answered correctly and incorrectly. Afterwards, have them turn in this summary answer sheet.
Once the data have been collected, debrief each individual, and explain that this task is meant to investigate how the costs and benefits of cheating influence a person’s decision to be dishonest.
To analyze the data, calculate the number of correct answers and plot as a frequency distribution showing the percentage of participants by group.
Notice that individuals allowed to self-report their results indicated that they solved a significantly higher number of matrices compared to controls.
However, based on the overlap of the distributions, most participants in the experimental condition cheated just a little with marginal inflation, and only a small percentage cheated a lot.
This suggests that participants weigh two things in their decision to be dishonest: The first is the external reward of $10 per correct matrix, which results in their lying about the number of problems they solved.
The second is the internal reward of their moral self-image—still being able to view oneself as an honest person—that effectively caps this lie, and deters participants from claiming they solved the maximum number of problems.
Now that you’ve learned how the Adding-to-10 task can be used to explore the relationship between the extent of cheating and self-image, let’s take a look at other ways researchers are investigating dishonesty.
On one hand, some scientists are evaluating how moral priming—reminding an individual about honest concepts and actions—influences their willingness to cheat. Students were asked to sign an honor code contract, and afterwards were administered the Adding-to-10 task.
It was found that there was no significant difference between the number of matrices these experimental group participants noted they solved and data from control individuals.
This suggests that prompting individuals with their institution’s honor code reminds them to be moral, and decreases their tendency to cheat.
In addition, such reminders of honesty are being extended in other settings, like an office, where supplies often go missing. In this case, it’s been shown that if a mirror is placed in an area where individuals might be tempted to steal—such as in the supply closet—they are discouraged from taking home all the pens.
The idea is that, by observing their actions, the potential thief is made self-aware, and they realize that taking all of them might mean they’re not as honest as they’d like to believe.
You’ve just watched JoVE’s video on how the Adding-to-10 task can be used to better understand dishonesty. By now, you should know how to design the experiment, collect control and cheating data, interpret the results, and grasp how self-image—especially reminding an individual of their actions—could be used to discourage cheating in other applications.
Thanks for watching!