Using a Data-Driven Context Model to Support the Elicitation of Context-Aware Functionalities – A Controlled Experiment

Background: Context modeling to support the elicitation of context-aware functionalities has been overlooked due to its high complexity. To help overcome this, we have implemented a data-driven process that analyzes contextual data and generates data-driven context models. Objective: We aim at investigating to which extent a data-driven context model supports the identification of more complex contexts (i.e., contexts that combine several contextual elements) and unexpected context-aware functionalities. Method: We used a one factor with two treatments randomized design with 13 experienced software engineers. Given a specific system-supported user task, the participants were asked to come up with requirements that describe context-aware functionalities to improve the user task. Results: Use of the data-driven context model increased the average number of contextual elements used to describe requirements from 1.77 to 4.23. No participant from the control group was able to identify by themselves any of the contexts included in the model. All comparisons between groups had sufficient effect size and power. The participants regarded the data-driven context model as a useful tool to support the elicitation of context-aware functionalities. Conclusion: The data-driven context model has shown potential to support the identification of relevant contexts for given user tasks.