A brain computer interface (BCI) is a communication system that allows users to control computers or external devices by detecting and interpreting brain activities. The initial goals of BCI were to help severely disabled people, such as people with "locked-in" syndrome, to communicate with the outside world by interpreting their brain signals into corresponding external commands. Nowadays, state-of-the-art BCIs, especially using Electroencephalography (EEG), bring benefits to normal and healthy computer users in a way that enriches their experiences of everyday Human Computer Interaction (HCI). Although EEG may be used in the same manner of continuous control and communications, it has been extended to assist and measure the inner states of users in a more passive way. Because of this, a new categorization of BCI systems has been proposed, dividing BCI applications in general and EEG-based systems in specific into active, reactive, and passive BCI. This thesis focuses on how portable and commodity EEG headsets can benefit the majority of HCI users with their limited capabilities, in comparison to clinical and expensive headsets. Our investigations focus on active and passive EEG-based BCI systems. We first investigate about how to use task engagement as an additional input besides traditional input methods in the context of active BCI. We then move forward to passive use of BCT by using task engagement to evaluate an application while the user is taking part in an interaction. We further extend our investigation to Event-Related Potentials where in particular Error-Related Negativity is used to detect users' error awareness moments. We show that using EEG signals captured by Emotiv headsets, moments of users' error awareness (or Error Related Negativity - ERN) can be detected on a single trial basis. We then show that the classification rates are sufficient to benefit HCI in single user. Next, we show ERN patterns can be detected in observation tasks where it not only appears in the observers' EEG, but also shows an anticipation effect in collaborative settings. Based on the results, we propose different scenarios where task designers can employ these results to enhance interactive applications, combining with popular HCI settings and input methods.