On the human-likeness of computers: Building blocks of anthropomorphism

Recent advances in Artificial Intelligence have reignited the question of whether computers are humanlike. This paper reviews and integrates various literatures that differ in whether they define human-likeness in terms of a) computers’ outputs, b) the mechanisms by which outputs are generated, or c) human responses to computers. Moreover, the paper introduces BBA (Building Blocks of Anthropomorphism), a novel framework that captures perceived human-likeness of computers and subsequent human evaluations and responses. BBA comprises a list of technological features (building blocks) which feed into two dimensions (Intelligence, Sociality) and four facets (Cognitive Ability, Autonomy, Subjectivity, Cooperation) that elicit anthropomorphism. Moreover, across four propositions, BBA sketches how the building blocks affect the perceived human-likeness and subsequent human evaluations of computers. Crucially, BBA posits that anthropomorphism does not always elicit favorable responses, depending on factors such as source disclosure, descriptive expectations, and normative expectations about computer capabilities. The framework thus provides a conceptual umbrella for a broad range of empirical findings, including Turing test scenarios, algorithm aversion, or the uncanny valley effect. Conceptual, empirical, and practical issues are discussed.