Detecting affective covert user states with passive brain-computer interfaces

Brain-computer interfaces (BCIs) provide insight into ongoing cognitive and affective processes and are commonly used for direct control of human-machine systems (Vidal, 1973). Recently, a different type of BCI has emerged (Cutrell and Tan, 2008; and Zander et al., 2008), which instead focuses solely on the non-intrusive recognition of mental state elicited by a given primary human-machine interaction. These so-called passive BCIs (pBCIs) do, by their nature, not disturb the primary interaction, and thus allow for enhancement of human-machine systems with relatively low usage cost (Roetting et al., 2009; and Zander et al., 2008), especially in conjunction with gel-free sensors. Here, we apply pBCIs to detect cognitive processes containing covert user states, which are difficult to access with conventional exogenous measures. We present two variants of a task inspired by an erroneously adapting human-machine system, a scenario important in automated adaptation. In this context, we derive two related, yet complementary, applications of pBCIs. First, we show that pBCIs are capable of detecting a covert user state related to the perception of loss of control over a system. The detection is realized by exploiting non-stationarities induced by the loss of control. Second, we show that pBCIs can be used to detect a covert user state directly correlated to the user´s interpretation of erroneous actions of the machine. We then demonstrate the use of this information to enhance the interaction between the user and the machine, in an experiment outside the laboratory.