3.4 Summary
4.1.2 Sensor Data as Learning Content
According to Fleck (see section 3.3.1) and the CSRL model (see sec- tion 2.1.3), captured data can support reflection. The data become the learning content that provides new insights to a user. However, in contrast to traditional learning content (e.g., from a text book), the captured data about an experience do not provide insights or new concepts on its own. The data can be understood only in light of the experience in which the data were captured. Sellen and Whittaker [97] speak of “Synergy not
substitution” to emphasize that the capturing of data cannot replace the
memory and the analysis by a user. The captured data have to relate to the experience of the user to lead to insights. As a consequence, the two main requirements for reflection content are (a) that learners can relate their own experiences to these data and (b) that this process can lead to new insights. Insights can range from an unobservable change in attitude to a change in behavior [4].
CSRL should not be confused with technology that involves only behav- ior change, such as persuasive technology (see Section 3.3.2). Although, CSRL and persuasive technology both involve supporting and influencing cognitive processes by technology, outcomes of reflection only rarely re- sult in direct observable changes, but rather change the mindset of the reflecting person. Examples are the initial awareness of a problem, changes in attitude, or simply the insight that there is no real problem. This section analyzes the similarities and identifies how CSRL can benefit from the advances in ubiquitous computing and in particular from persuasive technology.
Many applications from persuasive technology (see Section 3.3.2), such as [122, 124], create a feedback loop to influence behavior. The user is
constantly provided with feedback regarding current or past behavior to influence future behavior in the desired direction. Such feedback loops are also the basis of many reflective learning theories (e.g., the Kolb cycle [5] or the model by Schön [6]). The model by Boud et al. [4] is not explicitly designed as a loop, but the process of re-evaluating experiences can and probably has to be repeated to learn continuously. As described in Section 2.1, these reflective learning processes target an incremental improvement and learning from new experiences. Similarly, the CSRL model, which serves as the background for this approach, forms a cycle.
Figure 4.1 depicts how sensors and visualizations of sensor data can support reflection and behavior change. The model, similar to the Kolb cycle [5] and the model by Schön [6], is structured in a feedback loop. The naming of three of the four phases is based on the model of Boud et al. [4] because the CSRL model [33] itself is based on this model: Experience,
Reflective Process and, Outcomes. The fourth phase, Behavior, is not an
explicit part of Boud’s model, but the required source of experiences to close the feedback loop, similar to the models by Kolb and Schön. In the cycle, behavior results in experiences. During the reflective process these experiences are revisited and analyzed to come to new outcomes. Some of these outcomes will translate into an actual change in behavior that can be analyzed in subsequent cycles.
Reflective Process Outcome Behavior Feedback Measurement
Content for Reflection
Each transition between phases may fail. Behavior or the impact of own behavior might not be perceived at all and, therefore, does not become part of the experience. An important and relevant experience might go unnoticed because of the amount of experiences that we make every day. As a result, this experience never becomes an item of the reflective process. A reflective process might not be fruitful if it is not creating outcomes and may then turn into rumination. Finally, an outcome that could lead to a change in behavior might be prevented by the circumstances. For instance, an employee who decided to stop working late might suddenly be confronted with a large number of urgent tasks.
Technology can help at various points to minimize error and facilitate successful reflection, influencing attitudes and potentially changing be- havior. This thesis focuses on the support of the reflective process by facilitating the transitions from the behavior to the reflective process stage. Further technologies can support the other parts of the cycle. For instance, a system can guide reflection with reflective questions or help to document and communicate outcomes.
The envisioned sensor support creates a second loop with two new components that support the cycle. Behavior is not only perceived by the individual but measured by sensors. In addition, the captured data or a summary has to be provided as feedback to enrich the perceived experience with additional data [203]. The combination of experience and feedback will influence the reflective process, the resulting outcomes, and finally the behavior. The change in behavior is experienced and sensed to create a comprehensive feedback to influence the reflective process. According to Boud et al. [4], the reflective process involves a “stepping back from
experience.” This step to obtain a more objective perspective on behavior
can be supported by additional data that provide a different and often more objective perspective.
In summary, a system that is intended to support the reflective process has to close a second feedback loop by (a) capturing data linked to relevant behavior and (b) providing these data as feedback that can enrich the experience itself. The reflective loop and the feedback loop have to be as closely intertwined as possible. This design task is challenging because the internal reflective process can be observed only by its visible or articulated outcomes.
reflection session (see Section 2.1.3).
• Data that provide a different perspective on experience can actually trigger a reflective process by inducing a cognitive dissonance [31]. • In some cases, the data might act as markers to indicate a particular
point in time, a particular location or process upon which it is worth reflecting. Although these data do not lead to insights in themselves, they can point to relevant data or experiences.
• Finally, additional data may be required during the reflection session for an in-depth analysis of an event.
The impact of the collected data can be increased by aggregating the data or sharing it with others. The aggregation over time, multiple employees, or events can provide new perspectives on experiences. The resulting high-level perspective can be useful to abstract from the specific event and generalize insights for an individual (e.g. reviewing trends in bio-signals) or collaborative level (e.g. reviewing team performance over one month). Sharing of data enables all forms of social facilitation of reflection (see Section 3.1.3). Users can benchmark their own results in relation to those of their colleagues, or reflect in collaboration with them. For instance, employees could compare performance measures of a specific task to those of a more experienced colleague or to the average performance of all employees. The benefits of collaborative reflection are described in more detail in [93]. However, sharing of data is not always possible, or desired, at the workplace. Especially at the workplace, privacy considerations must be taken into account. Furthermore, colleagues have to be able to understand and interpret the data from their own perspectives.