Improved Cognitive Control in Presence of Anthropomorphized Robots

Abstract

There is evidence that attentional control mechanisms in humans can be boosted in performance contexts involving the presence of other human agents, compared with isolation. This phenomenon was investigated here with the presence of artificial agents, that is, humanoid robots in the context of the well-known Stroop task requiring attentional control for successful performance. We expected and found beneficial effects of robotic presence (compared with isolation) on standard Stroop performance and response conflict resolution (a specific component of Stroop performance) exclusively when robotic presence triggered anthropomorphic inferences based on prior verbal interactions with the robot (a social robot condition contrasted with the presence of the same robot without any prior interactions). Participants’ anthropomorphic inferences about the social robot actually mediated its influence on attentional control, indicating the social nature of this influence. These findings provide further reasons to pay special attention to human–robot interactions and open new avenues of research in social robotics.

Appendices

Appendix 1

For each following proposition, please note your level of agreement on the scale by surrounding the corresponding number (1–7). 1 = Strongly disagree, 7 = Strongly agree. There are neither good nor bad answers, only your personal opinion interests us. This questionnaire and its answers will remain totally anonymous.

These Traits are associated to the robot:

Civilized	1	2	3	4	5	6	7
Refined	1	2	3	4	5	6	7
Moral	1	2	3	4	5	6	7
Rational	1	2	3	4	5	6	7
Mature	1	2	3	4	5	6	7
Lack of culture	1	2	3	4	5	6	7
Coarse	1	2	3	4	5	6	7
Amoral	1	2	3	4	5	6	7
Irrational	1	2	3	4	5	6	7
Childlike	1	2	3	4	5	6	7
Emotional	1	2	3	4	5	6	7
Warm	1	2	3	4	5	6	7
Cognitively open	1	2	3	4	5	6	7
Autonomous	1	2	3	4	5	6	7
Deep	1	2	3	4	5	6	7
Insensible	1	2	3	4	5	6	7
Cold	1	2	3	4	5	6	7
Psychorigid	1	2	3	4	5	6	7
Passive	1	2	3	4	5	6	7
Superficial	1	2	3	4	5	6	7

Appendix 2

See Table 2.

Appendix 3

3.1 Error Rates

We conducted the same repeated measure analysis as before (RTs) on error rates (see Table 2 in Appendix 2 for error rates). No interaction effects were found either on Session x Type of conflict x Performance context interaction, F(2, 115) = .20, p = .816, $\eta^{2}_{p}$ < .01, the Session × Type of conflict, F(2, 115) = .20, p = .659, $\eta^{2}_{p}$ < .01, the Session x Performance context, F(2, 115) = 1.78, p = .173, $\eta^{2}_{p}$ = .03 and the Type of conflict x Performance context interactions, F(2, 115) = .41, p = .667, $\eta^{2}_{p}$< .01. Only main effects were significant. Participants produced less errors in Session 2 than in Session 1, F(1, 115) = 16.60, p < .001, $\eta^{2}_{p}$ = .04. We also found a main effect of the Type of conflict F(3, 115) = 51.78, p < .001, $\eta^{2}_{p}$ = .31: While there was no difference between standard Stroop and response conflict, F(1, 117) = 1.22, p = .272, $\eta^{2}_{p}$= .01, and no difference between semantic and task conflicts, F(1, 117) = .10, p = .751, $\eta^{2}_{p}$< .01, the level of interference associated with standard Stroop and response conflict averaged was higher than the interference associated with semantic and task conflicts averaged, F(1, 117) = 83.84, p < .001, $\eta^{2}_{p}$= .42.