Pim Haselager - Did I do that? Brain-Computer Interfacing, Intelligent Devices and the sense of agency

In Brain-Computer Interfacing, brain activity is measured and used to control a device such as a wheelchair or a computer (Lebedev & Nicolelis, 2006; van Gerven et al, 2009). Portable, non-invasive brain measurement techniques like electroencephalography (EEG) and near-infrared spectroscopy (NIRS) are commonly used, but low signal to noise ratio’s, low processing speed, and huge intra- and inter-subject variability prevent a rapid deployment of clinical and consumer BCI applications. Providing the devices with artificial intelligence (e.g. environment-sensing, obstacle-avoidance and path-finding capabilities for intelligent wheelchairs) can significantly enhance the overall performance of the brain-machine combination (Vanacker et al. 2007). The combination of BCI with such intelligent devices (ID’s) is valuable for the rehabilitation and daily-life support of e.g. disabled patients.

However, delegating parts of the overall task to intelligent artificial controllers of the applications can affect the experience of oneself as causing or generating an action and may produce or increase uncertainty about a user’s ‘ownership of action’ (Wegner, 2002) or ‘sense of agency’ (Gallagher, 2000). Particularly confusing cases can arise when the artificial control is of a brief compensatory nature, e.g. taking over when brain signals temporarily become unreliable (because of noise) or ineffective (because of lapses of attention due to distraction or tiredness). Potential mismatches between user’s intentions and the sensed results are manifold: The user could have failed to produce the intended brain signal (making an error in e.g. an imagined movement task). An adequate brain signal could have been measured incorrectly, or classified incorrectly by the BCI. The ID could have taken over control improperly (e.g. on the basis of an unjustified rejection by the BCI of a brain signal or a mistaken measurement of distraction). Finally, the ID could have been justified in taking over control, but erroneous in the produced behavior. In such situations, whether successful or unsuccessful, users may be insecure about the extent to which the resulting behavior (e.g. the wheelchair heading for a particular direction) is genuinely their own. Though they may be certain about what they wanted, they may be insecure about whether they did or did not ‘do’ it. This presentation will examine such cases of insecurity and examine their potential implications for legal responsibility (Vincent et al., 2010).