Group Writing Exercise: What I Learned

This group-writing exercise proved to be very helpful for a number of reasons. First, the arrangement of the groups worked very well. Because there was (or was supposed to be) a member of the group who participated in my own experiment as well as two other people unfamiliar with the project, different perspectives offered different criticisms. Also, I learned that reading others’ papers can help me see what I want to or do not want to incorporate in my own paper. We were reading the papers to offer constructive criticism, but in the process, we were able to see strong sections of each other’s writing as well. I also got practice in how to give criticism in a positive way and receive criticism politely and without defense. As an author, the group-writing exercise greatly helps to put the finishing touches on my paper. As a reviewer, the exercise opens my eyes to others’ writing techniques and general ideas as well as my own abilities to look at others’ work with a critical eye.

1. Just to remind you our hypotheses, we predicted that reaction time would be slowest for the incongruent condition, faster for the neutral condition, and fastest for the incongruent condition. We also predicted that accuracy would be smallest for the incongruent conditions, larger for the neutral condition, and larger for the congruent condition.
2. We used the one-tailed, paired t-test for our statistical calculations. This test helps us to discern any differences between groups in a within-subjects design. We used a one-tailed test because we have a directional hypothesis (due to a priori notions of how human cognitive functioning works).
3. Because all subjects successfully completed all trials, we didn’t discard any data. Before actually conducting any tests, it was evident that the general pattern of results supported both of our hypotheses.
4. Here are the results for reaction time. Reaction time is shown in milliseconds. The congruent condition was the fastest, the neutral condition was slower, and the incongruent condition was the slowest. The error bars represent the standard error of the means, or the general interval where the mean could fall if we measured the entire population.
5. Each t-test did show significant difference between conditions.
6. Here are the t-values, all which surpassed the critical value of 1.717 for 22 degrees of freedom. The p value is the probability with which our results occurred by chance. A larger t value proposes a larger difference, and the largest t value was between incongruent and congruent conditions.
7. Here are the results for accuracy, expressed in percentages. The mean accuracy was largest for the congruent condition, smaller for the neutral condition, and smallest for the incongruent condition.
8. The t-tests, however, showed a significant difference only between the incongruent and congruent conditions as well as the incongruent and neutral conditions. While the mean accuracy percentage for the neutral condition was smaller than the congruent condition, the two are not statistically different based on the data from our experiment.
9. This is shown here. Between congruent and neutral conditions, the t-value is only .85, which does not surpass 1.717.

2. The statistical tests most appropriate for our data are paired t-tests. We chose these tests because we are comparing two sets of within-subjects data at a time to see if there are any differences between the groups. A t-test was conducted between the gain-framed questions and loss-framed questions as well as between the high-payoff-framed questions and the low-payoff-framed questions. For the interaction effect, a t-test was conducted between the mean difference of the answers for gain and loss-framed questions in the high payoff scenarios and the mean difference of the answers for gain and loss-framed questions in the low payoff scenarios. If the differences turn out to be statistically insignificant, the graphs will be parallel, suggesting no interaction (the level of payoff does not affect the gain or loss framed scenarios differently). Because our measure is a forced-choice survey and we, the experimenters, are entering the data into the computer, we do not expect errors or any other special case for which we need to discard or manipulate data.

3. In the survey, subjects answer 8 questions with one of two choices. One of the choices is a sure option (“this will happen”), and the other option is a risky option (“either this will happen or that will happen”). We have coded the answers from the subjects into an Excel spreadsheet. The answers to all 8 questions for each subject are recorded. Sure options are recorded as “1,” and risky options are recorded as “0.” This way, when performing t-tests, the means of the data per subject (or per type of question) will represent the total percentage of sure-option responses. (i.e. if subject A responds with a sure-option to question 1 and a risky option to answer 2, then the mean for those two questions will be .5). These percentages serve as our dependent variable.

4. Two-tailed, paired t-tests were implemented in each data analysis. Answers to loss-framed and gain framed-answers were compared; the mean percentage of sure-option responses for the gain-framed questions was .500 (standard deviation = 0, standard error =0). The mean percentage of sure options responses for the loss-framed questions was .333 (standard deviation = .144 and standard error = .083). Results are approaching significance with t(2) = 2.000 and p= 0.18. Concerning high-payoff and low-payoff scenarios, percentage of sure-option responses for the low-payoff questions was .333 (standard deviation = .144 and standard error = .083) and the percentage of sure options for the high-payoff questions was .417 (standard deviation = .144, standard error = .083). Results again approach significance with t(2) = .500 and p=.67. An interaction between payoff-type and gain/loss frame type is suggested; the mean difference between gain and loss frames for the low-payoff scenarios is .333 (standard deviation = .289, standard error = .167). The mean difference between gain and loss frames for the high-payoff scenarios is 0.00 (standard deviation = 0, standard error = 0). Results approach significance with t(2) = 2.000 and p=.18.

Masking Experiment Statistics Assignment

The appropriate test to use for this demonstration is the t-test for a two-group, within-subjects design experiment. If the null hypothesis is true, then the difference between the means of the masking and non-masking groups will be zero. I would typically calculate the estimated variance and estimated standard error of the means for my t-test, but in keeping with the Excel Assignment for this data, I will use the sample statistics instead of estimated population statistics. The independent t-test may also have seemed appropriate, but our data is from a within-subjects design, not a between-subjects design.

If the alternate hypothesis is true, then our test will support that there may be a difference between the two conditions. The criterion used in the decision will be a critical t value. This value implicates 1Q (because we assume that the masking variable will impede performance) and an alpha level of .05. The critical T value is thus, for 22 degrees of freedom, 1.717. The obtained t-value is 9.165. Because this value exceeds the critical value, we can reject the null hypothesis.

References

Gonzales, C., Dana, J., Koshino, H., & Just, M. (2005). The framing effect and risky

decisions: examining cognitive functions with fMRI. Journal of Economic

Psychology, 26, 1-20.

Kim, S., Goldstein, D., Hasher, L., & Zacks, R. T. (2005). Framing effects in younger

and older adults. Journal of Gerontology: Psychological Sciences, 60B(4), 215-

218.

Kuhberger, A., Schulte-Mecklenbeck, M., & Perner, J. (2002). Framing decisions:

hypothetical and real. Organizational Behavior and Human Decision Processes,

89, 1162-1175

While the results of this study expand upon previous research on the framing effect, more investigation will be needed to converge upon a theory for the effect in low and high payoff scenarios. The ability to generalize our results across populations is also key. Scenarios were devised with the subject pool in mind; high-payoff scenarios offered consequences that were thought to be especially relevant and important to Psychology 331 students and instructors. What results would our questionnaire elicit in a completely different population, and would a more generalized measure yield as strong results? In addition, current research suggests that the framing effect is more pronounced in older adults (Kim, et al. 2005), a variable that we were not able to concretely examine. Future investigation may include similar procedures across two (or more) distinct age groups.

Though we were interested in examining the framing effect as a whole, we gained particular insight into how the frame of a scenario affects human decision-making across relatively large and small (albeit hypothetical) consequences. This is an important examination that can lead to a number of real-world inferences. For example, perhaps larger and more encompassing problems involving political, social, consumer, and medical decisions incur safer choice that are produced by the frame of the situation at hand. In our experiment, not only was the tendency to choose sure options in gain-framed scenarios and risky options in loss-framed scenarios supported across subjects, but scenarios involving higher payoffs also elicited surer responses. The fact that neither previous studies nor our own data supports our last hypothesis (that the framing effect would be more pronounced in high-payoff situations) is of note. Further investigation and refinement of procedure will undoubtedly help uncover how our decisions are influenced not only by the frame of the background information but also the magnitude of the consequences.

As with any thoughtfully-executed experimental design, our procedure was not without limitation. All subjects completed the survey in the same room and at the same time as multiple subjects; the anxiety caused by wanting to finish before (or at the same time as) other subjects may have interrupted the cognitive processes used to make the decision in each scenario. The length of the survey may also have caused subjects to tire and pay less attention to the questions towards the end. Also, even though we offered a forced-choice survey, some subjects felt it necessary to provide written justification for some of their selected answers. Justification for a decision has been shown to weaken the framing effect (Kim, et al. 2005), and those subjects providing justification may have confounded the study. In addition, because the degree to which a scenario offered a low or a high payoff was determined by the experimenters and not subjected to measures of validity, the results of the experiment may not be as legitimate as possible.

Interestingly, our third hypothesis was unsupported. We proposed that the framing effect would be more pronounced in high-payoff scenarios as opposed to low-payoff scenarios. This would have been shown with an interaction between frame-type (gain versus loss) and payoff type (high versus low) with a larger discrepancy of sure-option responding between gain-framed and loss-framed scenarios in high-payoff scenarios. This was not the case; while there was an interaction, the discrepancy was actually much larger in the low-payoff scenarios and not at all significant in the high-payoff scenarios. This evidence is contradictory to previous findings. Kuhberger, et al. (2002) http://seattletimes.nwsource.com/html/sports/ that the framing effect was larger in scenarios offering a high payoff and proposed that the results were simply due to the magnitude of the consequences involved. (For example, both choices for a low-payoff situation seem equally small while in a high-payoff situation, the choices seem to offer distinctly different outcomes). Our data, however, did not follow this trend. Perhaps the subjects focused more intently on the high-payoff scenarios and were thus less susceptible to their frames than the general content of the scenarios themselves. Low-payoff scenarios, on the other hand, inherently offered consequences of less importance. Subjects may therefore have been less concerned with the actual content of the scenario and more susceptible to its frame.

As expected, the results support our first hypothesis: the framing effect exists and affects decision-making processes in humans. In our experiment, subjects responded significantly more often with sure options than with risky options when the scenario is framed as a gain. Alternately, subjects responded significantly more with risky options than with sure options when the scenario was framed as a loss. Our results represent a successful replication of multiple studies of the framing effect (Gonzales, et al., 2005; Kim, et al., 2005; Kuhberger, et al., 2002). A possible explanation for our results is the cognitive-affective tradeoff model proposed by Gonzales, et al. (2005): the cognitive effort required in the decision-making process and the feeling of pleasure elicited from the decision will affect what choice is made. For example, when subjects in our experiment read the scenarios and the answer choices, the risky option for gain-framed scenarios (disregarding payoff level) may have incurred greater cognitive processing or feelings of displeasure in the subjects, leading them to choose the sure choice. Gonzales and colleagues supported their model with brain imaging techniques; future replications of this experiment may also want to include fMRI scans to physiologically test the cognitive-affective tradeoff model as an explanation for results that ultimately support the framing effect. WCFL Informational Guide.

The results of our study also support our second hypothesis: people are less risky in scenarios involving high payoffs than they are in scenarios involving low payoffs. This is shown by subjects responding significantly more with sure-option responses in the high-payoff scenarios than in low-payoff scenarios. These results are congruent with previous findings; Kuhberger and colleagues (2002) report that for scenarios involving real payoffs, subjects more often chose the sure option when the payoff was high and the risky option often when the payoff was low. A reasonable explanation for this pattern is that in general, people are less likely to exhibit risky behavior when there is “more to lose.” Specifically, in our study, the thrill of choosing a risky option for Drumheller Duck, Superman, Spiderman, or a Martian colony may outweigh the confidence or pleasure elicited from choosing a sure option. For the scenarios in our questionnaire in which the decisions for the scenarios carried larger consequences, the sure option may have seemed more safe and appropriate. Because we only implemented hypothetical scenarios, our study suggests that Kuhberger, et al. (2002)’s results can be applied to scenarios like ours in which the consequences of the decisions are not actualized.

Dee dee dee

Procedure

Three administrations of the surveys were given: one to participants in section A, one in section B, and one in section C. In each class, all subjects were present in the testing room and were administered the surveys at the same time. A hard copy of the survey along with an answer sheet was placed face-down in front of each subject, and subjects were instructed neither to turn over the materials nor begin until instructions were given. The experimenter then verbally informed the participants that they would be answering a survey involving decision-making. Subjects were also told that the questions involved hypothetical scenarios but responses should be made as if each decision would incur real, actualized consequences. Subjects were instructed to write their selections to each question (either A or B) on the provided answer sheet. The experimenter monitored the subjects during the administration, and when it was apparent that all subjects had completed the survey, all materials were collected and subjects were thanked for their time.

Design

A 2 (general frame: gain, loss) x 2 (payoff level: high, low) within-subjects design was implemented. All subjects experienced all levels of each independent variable (general frame and payoff level). The gain-framed questions involved scenarios in which either choice selected by the subject would result in an acquisition. Likewise, choices selected for the loss-frame scenarios incurred a loss. The high-payoff questions included scenarios about Psychology 331 grading techniques, studying habits, a worldwide flu epidemic, and Hurricane Katrina funds, which were chosen specifically because of their importance to Psychology 331 students. Low-payoff questions, on the other hand, involved scenarios with a hungry duck, a Martian mining program, Superman, and Spiderman. These questions were devised to offer relatively smaller consequences. The dependent variable was the percentage of total sure-option responses for the gain-framed and loss-framed questions as well as for the low-payoff and high-payoff questions.

Method

Subjects/Participants

Subjects consisted of 44 participants (11 males and 33 females) who enrolled in Psychology 331, sections A, B and C, at the University of Washington. Subjects consented to participation at the time of enrollment. Of those who participated, 41 were students in the class and three were instructors.

Materials

The only measure used in the experiment was a questionnaire devised by the experimenters. The questionnaire consisted of eight questions; each question contained a background scenario for which the subject was to make a choice affecting its outcome. The forced-choice design of the measure offered subjects either a sure answer (where the outcome has a 100% chance of occurring) or a risky answer (where one of two outcomes will occur and the probability of each actually occurring is given). Four questions were framed as a gain, and four questions were framed as a loss. For each of these frames, two questions offered a high payoff (relatively large consequences) and two questions offered a low payoff (relatively small consequences).

Bearing the aforementioned research in mind, we propose three general hypotheses concerning scenario-frame and payoff-type only. First, the framing effect is a cognitive phenomenon and does exist. To replicate previous findings, this will be shown with a greater total percentage of sure-option responses to gain-framed scenarios and a smaller percentage of such responses to loss-framed scenarios. Second, we expect that people will be generally less risky when situations involve large consequences (or a high payoff) and relatively less risky when the consequences are small (involving a low payoff). This will be shown with significantly greater total sure-option responding in high-payoff scenarios than in low-payoff scenarios. Third, we propose that the framing effect will be more pronounced in high-payoff scenarios than in low-payoff scenarios. This will be shown with an interaction effect between scenario-frame and payoff-type, with a larger discrepancy of sure-option responding between gain-framed and loss-framed scenarios in the high-payoff scenarios and a smaller discrepancy in the low-payoff scenarios. All hypotheses will be tested with a forced-choice questionnaire in which subjects will choose one of two decisions (a sure option or a risky option) scenarios involving either a gain-frame or a loss-frame with either a high or a low payoff.

A study of particular interest involves two more variables in addition to the gain versus loss scenario-frame: hypothetical versus real scenarios and scenarios with a high versus large payoff. In their experiment, Kuhberger, Schulte-Mecklenbeck, and Perner (2002) manipulate loss and gain-framed scenarios in terms of their actual payoff (subjects either are or are not exposed to the consequences of their decisions) and the magnitude of the consequences of their actions (“payoff,” in other words, is either relatively large or small). Results of the experiment suggest that subjects tended to be riskier in small-payoff conditions as opposed to large-payoff conditions, and the framing effect was generally equal across conditions between hypothetical and real scenarios. Because subjects are instructed to make their best decision to each scenario, regardless of whether the consequences are real or hypothetical, the cognitive effort employed and feelings of pleasure elicited from each decision may be equal across real and hypothetical conditions. Also, because the high-payoff scenarios offer consequences of relatively greater importance, subjects may not be as willing to take risks; more is at stake.

In relatively new framing-effect literature, evidence is shown that older adults are more likely to be susceptible to the framing effect than younger adults (Kim, Goldstein, Hasher & Zacks, 2005). This may be due to the likelihood that older adults will rely more on heuristic processing: general strategies that typically produce a correct solution to a problem. Heuristic processing conserves cognitive effort and, in turn, may greatly influence the extent to which framing influences decisions. When subjects were asked to provide justification for their decisions, however, the framing effect was equally pronounced across the age span (perhaps because justification is the product of more systematic processing and therefore lessens subjects’ own vulnerability to the framing effect). Because our study involves only adult participants, these findings are of importance.

A number of theories have been presented to explain the framing effect; present research suggests a cognitive-affective tradeoff model (Gonzales et al., 2005). In this model, the framing effect is explained by the cognitive efforts required to make a decision (greater cognitive effort makes the decision less desirable) and the emotion that the decision brings to the decision-maker (choices that bring feelings of displeasure are less desirable). Results of the study support the model, and it serves as a general framework to explain the effects that framing has on our decision-making abilities.

The Framing Effect: An Examination Involving High-Payoff and Low-Payoff Scenarios

The framing effect is a phenomenon of interest in the domain of human cognition. Its main premise is that decision-making is affected by the language in which the background information presented. Gonzales, Dana, Koshino, and Just (2005) summarize the effect well: when decision-making scenarios are framed as a gain, people tend to select sure-option decisions; when those scenarios are framed as a loss, people tend to select more risky decisions. This phenomenon occurs even when the expected values are equal. For example, imagine a problem in which someone is told that an emerging flu strain is expected to kill 6,000 people. The person is given descriptions of two programs to combat the flu and is asked to express which program he favors. In Program 1, 2,000 people are expected to be saved. In Program 2, there is a one-third probability that everyone (6,000 people) will be saved and a two-thirds probability that no one will be saved. Because the programs are described in terms of lives saved (gain-framed) instead of lives lost (loss-framed), the person is more likely to select the sure option (or Program 1, where the proposed option has a 100% chance of occurring) than the risky option (or Program 2, where probabilities of occurrence are assigned to either of two outcomes).

Abstract

The framing effect is a phenomenon by which decision-making in humans is influenced by the language or context in which background information is presented. In our experiment, we sought to examine the influence of payoff size (or magnitude of the consequences as determined by the background information) on the framing effect. We proposed three following: 1) the framing effect does exist, 2) people tend to be more risk-seeking in low-payoff scenarios, and 3) the framing effect will be more pronounced in high-payoff scenarios. To test our hypotheses, we used an eight-question, forced-choice questionnaire that presented subjects with both gain-framed and loss-framed scenarios with high-payoff and low-payoff conditions. The results of the experiment support the first two hypotheses, but not the third; the framing effect was actually more pronounced in the low-payoff scenarios. From these results, we can conclude that the framing effect does exist and that payoff size exerts a significant influence on the framing effect.

References

Gonzales, C., Dana, J., Koshino, H., & Just, M. (2005). The framing effect and risky

decisions: examining cognitive functions with fMRI. Journal of Economic

Psychology, 26, 1-20.

Kim, S., Goldstein, D., Hasher, L., & Zacks, R. T. (2005). Framing effects in younger

and older adults. Journal of Gerontology: Psychological Sciences, 60B(4), 215-

218.

Kuhberger, A., Schulte-Mecklenbeck, M., & Perner, J. (2002). Framing decisions:

hypothetical and real. Organizational Behavior and Human Decision Processes,

89, 1162-1175