Hypothetical vs RealTime Effort Discounting of Extra Credit

  • Slides: 1
Download presentation
Hypothetical vs. Real-Time Effort Discounting of Extra Credit David W. Dempsey, Heidi L. Dempsey,

Hypothetical vs. Real-Time Effort Discounting of Extra Credit David W. Dempsey, Heidi L. Dempsey, Aaron Garrett, David Thornton, Seth Martin, Morgan Whetstone, Elizabeth Ussery Jacksonville State University There have been a few studies to date which have looked at the discounting of effort in an educational setting. Most notably, Silva and Gross (2004) developed a procedure where students were asked at the beginning of the semester to determine how many extra credit assignments they wanted to complete in order to earn extra points. The researchers found that there was a significant positive correlation between students’ grades and the number of assignments they chose to complete. Further, Kirby, Winston and Santiesteban (2005) found that students’ temporal discounting of money was correlated with college GPA, even after controlling for SAT scores. Williams and colleagues (2007) developed a computerized procedure to examine the rate at which students discounted extra credit points. They found that the majority of students did not discount at all and those who did had rather shallow discounting curves. They also found that discounting rates did correlate with overall quiz score percentage, but not overall course grade. This study along with the Silva and Gross (2004) study indicate a modest relationship between discounting of extra credit and course performance, but neither directly examined how student’s choices were affected by effort. Williams, Dempsey, and Dempsey (2008) made a second revision to their procedure to examine effort more directly by lengthening the delays, counterbalancing the presentation order, and looking for evidence of a magnitude effect. Students were informed that the researchers were interested in the amount of time they would be willing to “study” (operationally defined as reading a specified psychology textbook out loud) in order to earn varying amounts of extra credit points. The results indicated that a discounting curve did fit the data, but not exceptionally well. Furthermore, a magnitude effect was observed where the smallest amount of extra credit (0. 5 points) was discounted more steeply than the higher amounts of extra credit (2 and 4 points). The current study is an extension of the previous two studies by Williams and colleagues. In this study we used a new computer program (Dempsey, Garrett, & Thornton, 2008), which allowed us to change commodities (e. g. , discounting of money vs. points) and change the method by which discounting is assessed (e. g. , hypothetical choices or real time choices). Finally, we can change how the points are distributed over time (in a linear manner vs. following a standard learning curve). Method Participants A total of 134 general psychology students from a regional university in the Southeast participated in at least the online portion of the data collection (67 of these students also participated in the laboratory portion of the study). There were 50 males (37. 3%) and 84 females (62. 7%). The majority of the participants were White (56%), followed by African American (40%), with the remainder (4%) being from other minority or mixed racial groups. The majority were freshman (68%), followed by sophomores (23%), then juniors and seniors (9%). Most students lived on campus (50%) or commute less than 15 minutes to school (26. 1%). Students ranged in age from 17 to 58 years old with a mean age of 21 years and a median age of 19 years (SD = 6. 61). Procedure After completing the consent, students were taken individually into a room with a computer and asked to put on a headset with a microphone. Students were then logged into a computer program developed by the first four authors (Dempsey, Garrett, & Thornton, 2008; Dempsey et al. , 2008). After filling in the demographic information, students completed several discounting tasks which included temporal discounting of money, effort discounting of extra credit (the majority of scenarios were hypothetical in nature, but two had consequences), effort discounting of hypothetical exam grades, and effort discounting of real time extra credit scenarios. Hypothetical Extra Credit Scenario: Imagine that your teacher was offering you a **1 point** extra credit opportunity where you could choose to read an English literature book out loud for **1 minute** in order to receive the full amount of extra credit points, or you could choose instead to read for a shorter amount of time and still earn some percentage of the total number of extra credit points. Finally, you could choose not to read at all and earn 0 points now. How long would you choose to read to earn extra credit points? The points/time combinations were as follows: 0 points in 0 seconds, . 10 in 6 seconds, . 20 in 12, . 30 in 18, . 40 in 24, . 50 in 30, . 60 in 36, . 70 in 42, . 80 in 48, . 90 in 54, and 1. 00 in 1 minute. After participants made their choice, they were immediately moved into a consequence condition where they were given the opportunity to actually read for 1 minute to earn 1 point (see below). After this the majority of choices were hypothetical in nature and involved decisions about reading for varying amounts of extra credit (1 point, 3 points, 5 points, 10 points) across varying amounts of time (5 minutes, 10 minutes, 15 minutes, 30 minutes, 45 minutes, 1 hour, 1. 5 hours, and 3 hours). One additional scenario had a consequence (3 points in 5 minutes) which was embedded within the hypothetical extra credit questions. All point/time combinations involved the same linear scaling where students would earn the same percentage of the total points for the percentage of total time they said they would be willing to read. In both the consequence conditions and the real time scenarios, participants were presented with a screen containing a text reading box, which highlighted words as they read, and a ruler on the right hand side of the screen, which visually depicted how long they had been reading and how many points they had earned. The text that was placed in the text box comprised various chapters from a text file copy of Edith Wharton’s (1919) Ethan Frome downloaded from Project Gutenberg (http: //www. gutenberg. org/ebooks/4517). Further, in the real time reading scenarios the points were scaled in two different manners. The first was the traditional linear scaling, which is the same that was used with the hypothetical extra credit points. That is, they could receive the maximum number of extra credit points for reading the full amount of time (e. g. , 10 points in 10 minutes), or they could receive a percentage of the full points for reading the corresponding percentage of time (e. g. , 4 points in 4 minutes, 8 points in 8 minutes). In the secondition, the scaling of choices was based on the standard learning curve, such that even a little studying will result in substantially increased grades, whereas it takes a great deal of studying to receive the highest possible grade (e. g. , Ettlinger, 1926; Mazur & Hastie, 1978). In this scale more points were received at the beginning of the reading session (e. g. , 6. 37 points in 4 minutes) and the point distribution flattened out near the maximum (e. g. , 9. 23 points in 8 minutes). For the purposes of this study, the learning curve was modified so that the maximum reward was actually attainable, rather than serving as an unattainable upper limit (asymptote). From this point forward, this scaling will be referred to as “logarithmic” because the shape of the curve roughly approximates a logarithmic curve (increasing, concave down). Results Effect of Linear vs. Logarithmic Distribution of Points in the Real Time Reading Scenarios In order to calculate degree of discounting, the coordinates (reinforcer, delay) of each indifference point were converted to a proportion of the total amount or delay and were plotted on a graph to create an indifference curve, including a curve of the mean indifference points across participants. Then the area under the indifference curve (AUC) was calculated by the trapezoid method, as outlined by Myerson, Green, and Warusawitharana (2001). The graphs below depict the discounting of points/time in the real time scenarios. A repeated measures ANOVA with the 30 participants who completed all four conditions (1 point in 1 minute, 3 points in 5 minutes, linear 10 points in 10 minutes, and logarithmic 10 points in 10 minutes) showed that there was a significant difference in the proportion of time read, F (3, 87) = 12. 01, p <. 001. Pairwise comparisons revealed those participants in the 1 pt/1 min condition (M = . 80, SD = . 41) and in the 3 pt/5 min condition (M = . 78, SD = . 39) read significantly longer than either of the 10 pt/10 min conditions. Of the 10 pt/10 min conditions, those in the logarithmic scaling condition read significantly longer (M =. 57, SD =. 44) than those in the linear scaling (M =. 42, SD =. 47). This indicates that time matters—people read for longer when the maximum total lengths of time were shorter (1 and 5 minutes vs. 10 minutes)—and that scaling mattered in the longer lengths—people read for longer when the scaling was logarithmic rather than linear. When comparing the number of words read per minute there were not significant differences between the conditions (total numbers of words read are also depicted in the first set of graphs). Relationship Between Real Time Effort Discounting and Hypothetical Effort Discounting To compare hypothetical effort discounting with real time effort discounting, we conducted a repeated measures ANOVA comparing hypothetical points students chose in the 1 pt/1 min condition, the 3 pts/5 min condition, and the 10 pts/10 min condition with the points they actually earned in the real time conditions which mirrored the hypothetical choices. The ANOVA revealed that there was a significant difference between points chosen in the hypothetical situations (M = 3. 97) compared to points earned in the real time situations (M = 2. 92), F (1, 63) = 24. 16, p <. 001. To break this down further, three one-tailed pairwise t-tests were conducted to determine which groups were different from one another. People overestimated the number of points they would earn in the 1 pt/1 min condition, t (64) = 1. 80, p <. 05, the 3 pt/5 min condition, t (63) = 1. 80, p <. 05, and most notably in the 10 pt/10 min condition, t (64) = 5. 20, p <. 001. The means are found in Table 1. Thus, although previous research has found that there were no differences between the temporal discounting of real versus hypothetical money (e. g. , Johnson & Bickel, 2002; Madden et al. , 2003; Madden et al. , 2004), the current study demonstrates that when effort is involved, there is more of a disconnect between people’s hypothetical judgments and what they would actually do when asked to put forth the effort. Table 1. Comparison of Mean Points Chosen in Hypothetical versus Earned in Real Time Discounting Hypothetical vs. Real Discounting Hypothetical 1 pt/1 min Real 1 pt/1 min Hypothetical 3 pts/5 min Real 3 pts/5 min Hypothetical 10 pts/10 min Real 10 pts/10 min Linear Mean Std. Dev. . 91 (91%). 85 (85%) 2. 68 (89. 3%) 2. 50 (83. 3%) 8. 34 (83. 4%) 5. 36 (53. 6%) . 28. 36. 88 1. 05 3. 26 4. 60 Discussion Limitations and Future Research One of the obvious problems of this research project is that around half of the data from the logarithmic reading condition had to be excluded because of a miscoding in the computer program. Also, there were potential fatigue factors in play due to the fact that it took participants between and hour and an hour and a half to complete these measures. The real time reading conditions were counterbalanced to counteract this effect and analyses revealed that there were not order effects for either the hypothetical choices or real time readings. Another limitation is the fact that we cannot assess as much as we would like using the real time reading feature due to the fact that it simply takes too much time. Due to the lack of funds, we have not been able to compare discounting of real money with real time effort discounting. We have several other things in this data set that we need to analyze which will be the genesis for future research, such as hypothetical discounting in work situations, self-report measures of impulsiveness and procrastination, and other personality measures related to impulsiveness (e. g. , locus of control, need for cognition, academic delay of gratification). Finally, we need to conduct another study to determine if students’ hypothetical discounting of exam grades correlates with how much they actually study for a given exam. Poster presented at the ABAI Behavioral Economics Conference, March 2011