Year: 2022
Type: dissertation
Abstract: Interfaces are used to perform various tasks and in HCI particularly a human performs some task with the aid of a computer with the mediation of an interface. Tasks and interactions can be modelled as graphs, where the graph attributes contain information relevant to the understanding of the interaction task. Based on the interaction graph, it is possible to compute numerical task complexity measures that help compare the complexities of different tasks. However, determining the nature of tasks with manual evaluation is labor-intensive and does not work well with large-scale problems such as algorithmic design or evaluation of large datasets. In this work, we have shown that it is possible to algorithmically infer tasks structures from user interfaces and compute task complexity measures for the task structures represented by graphs. More specifically, the graphs contain descriptions of the components and the interaction modes associated with them, such as a tap. The graphs have been generated from Enrico dataset view hierarchies. The accuracy of the generated graphs is 53.5 % (90 % CI, 15 % ME). Majority of the errors are caused by issues in the underlying dataset. The computed task complexity measures include Wood’s task complexity and Halstead’s E measure. The task complexity measures behave in a fundamentally different way, and their applicability requires further validation. The results demonstrate that it is possible to computationally model and understand tasks performed by humans on interfaces based only on the interface structure. The ability to infer interface task structure as a graph and an adjacency matrix adds a novel perspective for analyzing and modeling user interfaces.
Source: Aaltodoc (Aalto University)
Author: Nieminen, Sami
Institution: Aalto University
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7124228375
