Foregrounding Algorithms: Preparing Users for Co-design with Sensitizing Activities

Abstract

Algorithms are present in many of our everyday activities. However, there is generally low awareness of their presence among users, and there are various conceptualizations to define them. Additionally, algorithms are often both complex and opaque. These characteristics raise challenges when applying co-design activities to the interaction design of algorithms. We argue that researchers can overcome these challenges by developing sensitizing activities: activities that foreground the presence of algorithms, thus raising algorithmic awareness and a shared understanding, without influencing their initial experiences and expectations. We share how we applied sensitizing activities in two case studies: sensitizing interviews, and diary studies together with two-phase workshops. We share our experiences applying these techniques to overcome the challenges of low algorithmic awareness and multiple algorithmic understandings of participants. Finally, we offer recommendations for researchers and practitioners when applying sensitizing activities in this design context and invite further methodological discussion on this challenging topic.CCS Concepts: • Human-centered computing → Interaction design; HCI design and evaluation methods; Participatory design.

Towards Tangible Algorithms: Exploring the Experiences of Tangible Interactions with Movie Recommender Algorithms

“…Consequently, as expressed in previous work [4], it seems this approach could also be applied to explore design opportunities to improve the interaction with algorithmic systems. In particular, this active participation of users will inform the design and ideation of TUIs meant to interact with recommendation algorithms, a scope not explored before.…”

“…As a previous methodological paper suggests [4], there are at least two difficulties with co-design exercises for algorithmic interfaces: first, a high chance of low algorithmic awareness; and second, difficulties for users to share a unified understanding of algorithms. Following the suggestions of these authors, researchers can apply sensitizing activities to reduce possible complications related to unawareness about algorithms and uneven understandings of the topic among participants of co-design exercises.…”

“…Previous research has not yet explored TUIs to interact with recommendation systems to the best of our knowledge. Additionally, considering the profoundly abstract notion of algorithms [4], we anticipated that participants could have an enormous challenge trying to imagine and suggest tangible solutions for the interaction with algorithms without an initial inspiration. Therefore, we decided to initialize our design exploration with preliminary design space.…”

“…After defining the preliminary design space, we recruited active movie recommender system users via word-of-mouth, mailing lists, Facebook, and WhatsApp groups of the university. Following previous methodological suggestions for this context [4], all invitations avoided any explicit mention of "movie recommender systems" or "algorithms" to prevent overly critical participants towards algorithms or similar topics. Additionally, the invitation explicitly invited users without a background in computer science, data science, or related fields to avoid possible biases related to a high level of technical knowledge about algorithmic recommendations or AI.…”

“…Through thematic analysis, we uncovered various design considerations expressed by users to support tangible interactions with personalized movie recommendations: (1) Distinctions between TUIs and GUIs; (2) TUIs replacing predominant interfaces; (3) Preference for single-device TUIs; (4) The relevance of granular control for TUIs; (5) Apparent transparency limitations of TUIs; (6) TUIs and algorithmic social computing; and (7) Overview of specific design choices, including advantages and disadvantages of soft, hard, rounded, cubic, and humanoid forms for this context. Based on these findings, we present Recffy: the first functional TUI designed to enhance awareness and control of an algorithmic movie recommendation system.…”