Predicting group benefits in joint multiple object tracking

Abstract

In daily life, humans often perform visual tasks together such as, for instance, solving puzzles or looking for a misplaced key. In such tasks, humans tend to distribute task demands, enabling them to reach a higher performance compared to performing the same task alone – a group benefit. Research on group benefits in joint visual tasks has found that the availability of performance feedback and information about each other’s actions are two important factors influencing group benefits. Yet, no study has systematically investigated for these two factors which information is minimally required for co-actors to 1) attain group benefits, 2) devise collaborative strategies, and 3) further extend an already attained group benefit. Across eight experiments, we systematically varied the information co-actors received in a joint multiple object tracking task along an information continuum to address these three points. We find that co-actors 1) already attained group benefits without receiving any information, 2) minimally required team performance feedback or information about the co-actor’s actions to collaborate, and 3) devised collaborative strategies more quickly with additional information. Furthermore, we used a linear regression to predict group benefits and find that several factors (including personality traits) explain non-redundant variance. Collectively, our findings suggest that group benefits in joint visuospatial tasks come for "free" as no information exchange between co-actors is required to attain them. To further boost group performance, co-actors make use of coordination mechanisms to devise labour division strategies and these mechanisms depend on exchanging action information and/or receiving team performance feedback.

Humans share task load with a computer partner if (they believe that) it acts human-like

“…These results also suggest that the description of an artificial partner does not only alter how participants co-represent the task of an artificial partner in a joint stimulus-response compatibility task (Stenzel et al, 2012) but critically, it affects humans' willingness to share task load in a joint visuospatial task. Moreover, our findings suggest that the left / right labour division that has been preferred by human-human pairs across several studies Wahn and Kingstone, 2020;Wahn et al, 2020b) is also readily adopted when humans and a computer partner share task load.…”

“…The 64 participants were evenly split across our four conditions. We chose 16 participants as our sample size for each condition to match the statistical power to our earlier study, in which we collected 16 pairs for each condition as well (Wahn et al, , 2020b. We also ran a power analysis (Power: 0.80; alpha = 0.05) using G*Power (Faul et al, 2007) and found that our sample size is sufficient to detect effect sizes smaller (Cohen's d = 0.65) than in our earlier study (all Cohen's d were larger than 0.85) for the same analyses that we perform in the present study.…”

“…In such tasks, humans often decide to share the task load to facilitate performance (e.g., one person looks for their friend in the left half of the crowd while the other person searches the right half). Indeed, research on joint visual tasks has found that humans devise division of labour strategies when performing, for instance, a joint visual search task (Brennan et al, 2008;Wahn et al, 2020a) or a joint multiple object tracking (MOT) task (Wahn et al, , 2020b) -for a recent review, see Wahn et al (2018). In the latter MOT study, pairs of participants were required to track the movements of target objects among distractor objects on a computer screen.…”