Joint Action

---

Joint Action is a common behavior which normally consists in the manipulation of a single object by more than one subject, including a shared goal and the aim of reaching a group benefit, that is outperforming the result an individual could reach alone. The range of these actions reaches from very simple ones like carrying a table together to complex ones such as controlling an airplane.

Distributed vs. redundant control

A general distinction can be made between distributed and redundant control in joint action, which is whether the actors are limited to certain contributing actions, often by external factors (distributed control), or the possibility to share responsibilities at their own disposal (redundant control).

Examples: driving a boat together, whilst one person is limited to using the paddles only for directing and the other one only for getting ahead could be an example for a distributed control, whereas having both actors being able to do both and coordinating actions via communication in order to get the best result out of it would rather be recognized as redundant control.

Contrarily, collaborations like for example preparing a meal together, whether one individual is chopping and the other one is frying or in a different mode, would not be considered as joint action at all, because these processes simply represent a division of labour, which shares the property of a common goal but does not have to be necessarily limited to the joint control of an object.

Distributed control

In both distributed and redundant control tasks one of the general questions is, whether a group benefit can be realized, or in other terms, how much the benefits of controlling an object together outweigh the costs of necessary coordination of actions.

Besides some distinctive cases mentioned below, in most of the available studies group benefits are not observable in joint action tasks with distributed control, that is individuals generally outperform dyads in these tasks.

The success of dyads performance in these tasks seems to depend on

- availability of multichannel-information about co-actor’s actions (exceeding visible effects on the jointly moved object),

- specific task demands (e. g. fine or coarse control),

- the degree of coordination required,

- co-actor’s interindividual skill differences.

Availability of multichannel-information about co-actor’s actions

The access to information about the co-actor’s actions seems to play an important role in realizing group benefits. G. Knoblich and J. S. Jordan demonstrated this by instructing research participants to track the movements of an object on a screen by moving a cursor jointly, whereas one participant could affect it only in terms of “up and down” and the other one in “left and right”-direction by pressing keys. In a second condition, the actors heard a tone whenever their co-actor pressed a button. In the third condition one individual simply had the full control over the cursor and completed the task alone. In the beginning, individual's performance exceeded dyads performance, whereas dyads reached but never exceeded individual's level after some time. Notably, this improvement only took place when participants were informed about their co-actor’s acting by hearing a tone. A separation between own and foreign action only by visual feedback however was not sufficient.

Similar could be observed, when participants got haptic instead of auditory information about their co-actors activities. For that purpose, van der Wel et al. let individuals and dyads tracking a target via a swinging pole (like a pendulum) by pulling a rope for each direction either bimanually or with control distributed between two actors. Also in this case the haptic feedback via the antagonistic pullings seemed to be crucial for dyads performance and they didn’t exceed individuals performances.

Specific task demands

Other findings suggest, that dyads can outperform individuals in distributed control tasks when special necessities are given, for example in terms of fine versus coarse control. In a cursor-task of Wahn et al. participants had to place a cursor very precisely on a target after tracking it, either individually with bimanual control or in dyads with separated control of horizontal and vertical position. In the approach-/tracking-phase individuals performed better but when it came to the fine tuning of the cursor's position over the target, dyads could realize a significant group benefit. Notably, in this case no other than the visual feedback about own and strange actions was provided. Thus, the group benefit in joint control tasks seems to depend on the specific task demands also.

Degree of coordination required

In this context, Wahn et al. suggest, that a major contributing factor to reach a group benefit could be the extent of necessary coordination needed in order to complete a task. Tasks which involve two spatial dimensions with distributed control between participants (for example classical cursor-tasks) need a lower degree of interpersonal coordination than tasks with only one spatial dimension involved, as in these cases own actions can be in direct conflict with the co-actor’s ones in terms of responsibility for opposite movements along one axis (for example a “pendulum”-task). Thus, a lower demand of coordination could contribute to an explanation why group benefits are more likely to be realized in distributed control tasks which involve two separated spatial dimensions.

Co-actor’s interindividual skill differences

Finally, differences in the abilities of the co-actors appear to be influential as well. In a recent study, Mojtahedi et al. let dyads of participants lift and balance an object physically, whereas every participant had to take responsibility for one of two handles. The outcome was compared to participant's individual bimanual performance with the same object. Basically, only “worse” participants benefited from joint action, whereas “better” ones tended to decrease in their performance. Thus, besides an unequal benefit from joint action, it can be generally stated that similarity in co-actor’s individual skills seems to predict a group benefit.

Redundant control

Joint action tasks with redundant control seem to be more likely to produce group benefits as well as an improvement of individual performance.

According experimental designs reveal that in such tasks

- dyads outperform individuals most often,

- training in dyads improves individual performance already after single-sessions,

- the control of objects is mostly distributed by free will, although this is not set a priori.

The experimental settings of Reed et al. and Masumoto & Inui provided relevant insight in this context. Reed et al. let participants lift an object towards a target position and subsequently decelerating it on target, whereas Masumoto & Inui’s subjects had to imitate a target force which varied periodically over time by continuously pressing force transducers whereas the target force as well as the produced force were shown on a screen. In both the initial experiments and the follow-up-studies participants decided to use their options by distributing the control, although that was not mandatorily set a priori. That is, that in the lifting-deceleration-experiment most often one subject became responsible for lifting and the other one for deceleration and in the force-transduction-setting, that participants reacted immediately with lowering their effort when the co-actor’s increased it and vice versa.

Interestingly, this “distributed control strategy in a redundant control setting” led to significant group benefits. Notably, these benefits did not show up when the same task was performed while participants believed to have a real partner but acting with a playback of human behaviour. Thus, a real online interaction seems to be necessary to realize such benefits.

Moreover, another set of studies revealed significant benefits of a previous dyadic training for individual performance in a follow-up. In some target-tracking tasks they found, that the improvement in individuals performance was significantly higher after they had trained with a “real” partner than after training alone, with a computer or with a playback of human action.

Comparison of distributed and redundant control

Taken together, individuals as well as groups seem to profit from joint action predominantly in a redundant control-manner. Whereas the outcome is often an interaction with distributed control as well, a crucial factor seems to be the opportunity to distribute control in accordance with different coordination strategies and individual capabilities, not due to fixed external restrictions.

Prospective research

Firstly, some questions investigated in distributed control tasks were not addressed in terms of redundant control so far, for example the effects of necessary degrees of coordination, types of control (fine versus coarse control) and interindividual skill differences.

Furthermore, the factor group size has seemingly not been addressed so far in both distributed and redundant control tasks. It could be tested, whether there is a critical number of contributors, which doesn’t promise further group benefits when overcame. Also the quality and nature of relations between co-actors could be taken more into account. In this context, a recent study of joint visual search revealed better performance of friends than of strangers. Also the “cost” of adaptation when switching co-actor’s could be focused by future research.

A methodological topic would be the introduction of more informative measurements of performance. Most often performance reflects just the degree of deviation between an individual and a joint performance in a quantitative manner but does not contain information about the quality of actual collaboration. Recent studies use a more sophisticated approach by comparing the values of a “simulated” joint performance, based on the measures of two individuals performances, and the outcome of the actual joint performance. The rationale behind is, that the difference in the simulated and actual values reflects the degree of true collaboration and sheds light on its character. Future studies could even go a step further – beyond the comparison of mere performance levels of any origin – and implement new approaches to compare different styles of collaboration and thus provide valuable insight in how group benefits come true.

Finally, ecological validity has to be taken more into account in prospective research. There is a huge span between the heavily isolated variables in joint action experiments and the everyday-life-issues they shall shed light on, namely between a joint cursor-control-task and the interaction among co-pilots jointly controlling an airplane, for example. Future research must find a way to build a bridge between these extremes.

References

This article "Joint Action" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Joint Action. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.

---