Figure 1
The Cozmo robot and the Simon Says game: Cozmo (used for one-time and repeated interaction conditions) is a tank-like robot with a screen-like face and blue squares as eyes. It can express a wide range of emotions using its eyes and – in addition – sounds. More information on the robot Cozmo can be found here: www.digitaldreamlabs.com/pages/cozmo. The Simon Says game (used for the no social interaction condition) is an electronic game that is equipped with pressable buttons that light up. When playing Simon Says, players see a series of buttons light up that are associated with a tone in a specific sequence that the player needs to mimic. After the player successfully mimics the sequence, the sequence would increase in length. The Simon Says game was chosen due to its resemblance to the Quick-tap game.
Figure 2
Overview of the three interaction conditions: Participants in the no interaction condition arrived to the lab and played the Simon Says game before completing the gambling task. Participants in the one-time interaction arrived to the lab and interacted with Cozmo once before completing the gambling task. Participants in the repeated interaction condition came in at the beginning of the week and were given a Cozmo robot to take home. They were instructed to interact with Cozmo at least 20 minutes per day. No other restrictions were made on their interaction. On the fifth day, they came back to the lab and then completed the gambling task.
Figure 3
Proportion of games played: The graph illustrates participants’ reported proportion of games across all participants. The majority of the games played included “Quick tap”, “Memory match” and “Keep away”.
Table 1
Average reported duration of interaction per day.
| DAY | AVERAGE DURATION | SD |
|---|---|---|
| 1 | 39.3 | 14.0 |
| 2 | 35.2 | 7.5 |
| 3 | 36.8 | 9.9 |
| 4 | 34.8 | 7.6 |
| 5 | 34.5 | 4.9 |
[i] Note: Duration reported is in minutes.
Figure 4
Trial sequence of the gambling task: At the beginning of each trial, a black fixation cross was presented centrally for 500 ms, followed by the presentation of two differently colored squares - left and right of the fixation cross. After another 500 ms, the fixation cross changed its color from black to white, which indicated to participants that they could now make their choice by either pressing “2” for the left square or “8” for the right square. After participants made their selection, the fixation cross was presented again for a jittered time interval of 400 to 600 ms. Afterwards, participants received feedback if they lost (“LOSE”) or won (“WIN”) the trial. “Self” versus “Cozmo” was blocked in this experiment and was only shown once at the beginning of a block (not shown).
Table 2
Results of the log-log growth curve model.
| VARIABLE | B | SE | Z VALUE | P |
|---|---|---|---|---|
| Intercept | –.28 | .37 | –.75 | .45 |
| Growth Curve | .45 | .04 | 10.14 | <.001 |
| Condition: Repeated interaction vs. One-time interaction contrast | –.44 | .16 | –2.65 | <.01 |
| Condition: No interaction vs. One-time interaction contrast | –.34 | .17 | –1.95 | .051 |
| Recipient: Self vs. Cozmo contrast | .08 | .12 | .62 | .53 |
| Growth × Condition: Repeated interaction vs. One-time interaction contrast | .28 | .05 | 5.06 | <.001 |
| Growth × Condition: No interaction vs. One-time interaction contrast | .28 | .06 | 4.81 | <.001 |
| Growth × Recipient: Self vs. Cozmo contrast | –.03 | .05 | –.58 | .56 |
| Condition × Recipient: Repeated interaction vs. One-time interaction for Self vs. Cozmo contrast | .23 | .17 | 1.32 | .18 |
| Condition × Recipient: No-interaction vs. One-time interaction for Self vs. Cozmo contrast | –.08 | .19 | –.43 | .66 |
| Growth × Condition × Recipient: Repeated interaction vs. One-time interaction for Self vs. Cozmo contrast | –.17 | .07 | –2.18 | .02 |
| Growth × Condition × Recipient: No-interaction vs. One-time interaction for Self vs. Cozmo contrast | .01 | .08 | .2 | .83 |
[i] Note: Significance testing was based on 33039 Degrees of Freedom. The “b” denotes the variable estimate and the “SE” denotes the Standard Error of the estimate.
Figure 5
Learning rates as a function of condition: Growth curve models allow us to examine differences in learning the higher probability outcome between conditions. The figure illustrates the significant Growth X Condition interactions, illustrating that learning rates were significantly different between the No-interaction and the One-time Interaction condition and the One-time interaction vs. the repeated interaction condition. The differences were such that those in the No interaction learned the fastest, followed by those who had repeated interactions and then those who had a one-time interaction with Cozmo. The points in the graph illustrate the percentage of picking the higher probability outcome.
Figure 6
Learning rates as a function of condition and recipient: The growth curve model showed significant learning differences in the Growth x Recipient x Condition interaction such learning rates were faster for Self vs. Cozmo in the repeated interaction condition compared to the one-time interaction condition. However, no differences between Self vs. Cozmo were shown between the one-time interaction and the no interaction condition; points represent percentages of picking the higher probability outcome.