The paper is dedicated to automated and semi-automated engineering of human-machine interfaces through artificial evolution: the state-of-art, the notable problems, and the preposition of a potentially feasible evolutionary algorithm. The A/B Testing method is already widely used in web interaction design, but it utilizes a single measurable goal and therefore risks random undirected modifications in the interface `phenotype', since its `genotype' is not distinguished. Also, purely interactive assessment of the potentially defective solutions' fitness by real users is undesirable, so the neural networks-based Kansei Engineering (KE) method is increasingly used for fitness evaluation. We reason that KE can be extended to consider any kinds of requirements, not just the emotional aspect, and outline the relevant techniques. Further, as an interface abstract model corresponds to a genotype, we propose a dedicated domain ontology to represent the genome, i.e. the structure of the design space. Finally, we summarize the combination of several existing AI and design methods in an evolutionary algorithm that may be used for automated development and improvement of web user interfaces in the coming era of the Big Interaction.
No takes yet. Share an insight, caveat, or question.
Bakaev et al. (2016) studied this question.