Knowing Action Boundaries: Understanding and Mitigating Over-Estimation Bias in Reaching

The typical person can navigate their world day in and day out without making too many errors, and without giving such navigation much thought. Such continued success would indicate that people know, more or less, what they can do when presented with an environment. Several decades of research, however, have shown that this is not the case. When asked to estimate, for example, the extent of their arm’s reach, individuals will typically over-estimate the extent of their reach by 10%. While this figure fluctuates from between 5% to about 30% depending on the study, the presence of such an over-estimation bias is robust and consistent across the literature. The current work began by assessing some possible simple experimental causes of this over-estimation bias in reaching, and continued on to examine the extent to which this bias influences not just estimates of one’s ability to act, but the actions themselves. Finally, the current work assessed to what degree participants could learn to be more accurate in both their actions and their judgments of their abilities.
A set of two studies found that over-estimation bias does not result from two typical aspects of the methods used in reaching experiments; participants are not driven to over-estimate their reach by the connotations of effortful extension contained within the word ‘reach’, nor were their biases apparently drive by the method by which they maintained their upright posture during the experiments. Over-estimation bias proves to be a real and robust phenomenon.
A subsequent set of three studies found that actions, while they show a slight bias to over-estimate, are not as biased as judgments. Further, the repetition within these experiments (detailed in Chapter 5) facilitated learning in participants: both participants’ actions and judgments showed improved accuracy from the beginning to the end of their participation. The methods of these studies do not afford participants any visual, tactile, or external feedback about their performance, so, most interestingly, all of this improvement is apparently driven by proprioceptive feedback, coupled with the intermittent visual information of the target’s location on the table. Taken together, these studies indicate that perceptions are consistently biased to a greater degree than actions, but both can improve with practice but without continuous visual feedback.