This section performs data analysis comparing the questionnaire data measured between the no sensor, sensor, and visible card groups in the Poker Experiment. Statistical comparisons were performed using the Mann-Whitney test comparing variables between Sensors and No Sensors conditions that performed the same task. Additionally, paired Wilcoxon tests were conducted to look at differences between paired opponents. Effect sizes were calculated for significant values.
Table 4.2. Poker p-value summary: participants with sensors vs. paired opponents
| Charity Gains | Charity Loses | |
|---|---|---|
| Ethical | p=1.0 | p=.39 |
| Respectful | p=.59 | p=.96 |
| Uncomfortable | p=.33 | p=.41 |
| Hindrance | p=.25 | p=.30 |
| Immoral | p=.26 | p=.35 |
| Trustful | p=1.0 | p=.56 |
| Unfair | p=.28 | p=.20 |
| Sensors | p=1.0 | p=.65 |
| Performance | p=.52 | p=.14 |
Table 4.3. Poker p-value summary: disadvantaged participants using sensors vs. disadvantaged participants with No Sensors or Visible Card
| Charity Gains | Charity Loses | |
|---|---|---|
| Ethical | p=.82 | p=.45 |
| Respectful | p=.92 | p=.33 |
| Uncomfortable | p=.81 | p=.71 |
| Hindrance | p=.19 | p=.71 |
| Immoral | p=.18 | p=.55 |
| Trustful | p=.29 | p=.62 |
| Unfair | p=.82 | p=.45 |
| Sensors | p=.21 | p=.09 |
| Performance | p=.93 | p=.41 |
In the Charity Gains motivator, when a participant performed well a charity of their choice also benefited. There were no significant results from analysis of this motivator.
In the Charity Loses motivator, participants benefited at the expense of a charity. Participants were told "In addition, (charity) will win an amount of money. There were no significant results from analysis of this condition, but two interesting trends occurred.
Figure 4.17. With sensors that collect information about emotion vs. task for Charity Gains motivator
A trend occurred showing participants with sensors expressed a preference for situations "with sensors that collect information about emotion" when compared with participants without sensors. A Mann-Whitney test shows that the difference is almost significant (p-value = 0.0911). A Kruskal-Wallis test comparing this same preference across all three conditions (No Sensors, Sensors, and Visible Card) finds no significant difference (p-value = 0.1293).
Figure 4.19. Performance of disadvantaged players with sensors vs. their advantaged opponents for Charity Loses motivator
A trend occurred showing that participants who were in the sensor condition performed worse than their paired opponents. The mean of the disadvantaged players with sensors was 4.7 dollars won compared to a mean of 7.2 dollars won by their opponents. A Mann-Whitney test shows that the difference is not significant (p-value = 0.1443).
Performance was also compared among all "disadvantaged" subjects from different motivators but the difference was found to be non-significant.
A manipulation check comparing participant reports good or bad intentions in the different motivators failed; it was non-significant.
Additionally, some other analyses were performed to examine the disadvantaged participants in more depth. By grouping disadvantaged participants based upon their condition, additional Wilcoxon tests were performed.
We asked if people in one of the genuinely disadvantaged conditions (revealing a card or affect sensor data to their opponent) felt any different than their "disadvantaged" control, who just sat in the room where the sensors were located. There were no significant differences between these groups. However, there was a trend toward preferring sensors (p=.08) among those who had sensors or a card shown.
If instead a comparison is performed between disadvantaged participants who used sensors with participants having one of their cards shown to an opponent the following is observed. Participants having to show their card reported the situation to be significantly more unfair (p=.05) than those who had sensors.
These additional tests were performed after the original hypotheses the thesis set out to test were analyzed. Given that 54 individuals statistical tests were performed, it is entirely possible that the results reported in this chapter are entirely due to chance observation. Further discussion of these results occurs in Chapter 7.