Motor behaviour mimics the gaze response in establishing joint attention, but is moderated by individual differences in adopting the intentional stance towards a robot avatar

ABSTRACT

Leading another person’s gaze to establish joint attention facilitates social interaction. Previously it was found that we look back at agents who engage in joint attention quicker than at agents who display this behaviour less frequently. This paper serves to fill in two remaining knowledge gaps. Firstly, we examined whether this looking-back behaviour is replicated by a manual response. In an online task, participants selected one of two objects. One robot identity on the screen followed the selection most of the time, whilst the other looked at the other object most of the time. Participants clicked back on the following robot quicker when it looked at the same object relative to when it did not. We found no such difference for the unfollowing robot. Secondly, we examined how individual differences in autistic traits and adopting the intentional stance affected participants’ behaviour over time. The results showed that autistic traits influenced participants’ motor responses over time across conditions, whereas this response was only sensitive to the intentional stance for the unfollowing robot. Our results indicate that there is a potential overlap between evoked gaze behaviour and manual actions. However, individual differences may not fully predicate reflexive social behaviour.

KEYWORDS:

• Joint attention
• gaze-leading
• social robotics
• motor response
• growth-curve modelling

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data and scripts related to this manuscript can be retrieved on https://osf.io/brjqd/

Notes

1 Based on a similar pilot study, we did not expect this to be a relevant factor. 52 out of our final n = 80 participants self-reported using the mouse. No manipulation checks involving this variable were statistically significant.

2 Given the unbalanced number of trials in each condition (i.e. 80 / 20%), we corroborated our main trimmed reaction time analyses with the median reaction times as the dependent variable to justify reporting mean analyses subsequently. Main effect of following behaviour: F(1, 77) = 7.35, p = .008, η² = .003; main effect disposition: F(1, 77) = 0.43, p = .51; interaction effect F(1, 77) = 12.7, p < .001, η² = .005.

3 In exploration, we examined these analyses with accuracy error of landing coordinates relative to the centre of the face image. We found a main effect of following behaviour in which followed trials yielded greater accuracy than unfollowed trials, F(1, 77) = 10.1, p = .002, η² = .022. However, there were no other statistically significant main effects or interaction effects. Further, the trade-off between accuracy and reaction times is not straight-forward, as illustrated by the lack of correlation between reaction time and accuracy (r = .002, p = .99).

Additional information

Funding

This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant awarded to A.W., titled “InStance: Intentional Stance for Social Attunement” [grant number: 715058]).