Exploring a Behavioral Model of "Positive Friction" in Human-AI Interaction
Full paper available here: https://arxiv.org/abs/2402.09683
(This preprint has not undergone peer review or any post-submission corrections. The Version of Record of this contribution will be published in Springer Nature Computer Science book series in Volume HCI International 2024)
Figure 1. Mapping the relationship of positive friction and user behaviors in inhibition vs stimulation against resultant intentionality vs expansion of purpose.
Extended Abstract
There is a long-standing presumption in user experience design that friction is bad. This belief often leads to concerted efforts to minimize friction in user experiences, with the goal of making desirable actions easy and efficient. However, friction can be genuinely beneficial in certain situations, such as the insertion of deliberate delays to increase reflection in the form of waiting periods, preventing individuals from resorting to automatic behaviors, and enhancing opportunities for novelty and unexpected discoveries. More recently, the popularization and availability of AI on a widespread scale has only increased the need to examine how technology both reduces and introduces friction, and to what end. This suggests a need to investigate the various ways in which friction can be beneficial to user experiences, understand how user experiences with AI impact potential frictionlessness, and describe how future AI development may be productively informed by insights into their relationship.
Discussions of friction in designed user experiences is perhaps most familiar in the context of affordances (Gibson, 1977; Norman, 2013), or more recently in behavioral science’s emphasis on making good choices easy or fixing poor “choice architecture” (BIT, 2014; Thaler & Sunstein, 2008). Reducing UX/UI friction to reduce friction as a means to increase the efficiency and seamlessness of digital user experiences remains a common design strategy, as seen in Amazon’s One-click purchasing and Amazon One palm recognition shopping. In contrast, user experiences in which friction is intentionally introduced for deceptive purposes — such as UX/UI “dark patterns” (Gray et al, 2018) or behavioral notions like “sludge” (Sunstein. 2022)— present a clear and compelling reason for reducing friction.
However, friction can also be strategically employed to promote beneficial behaviors, and several design traditions within Human-Computer Interaction (HCI) challenge this prevailing focus on frictionless engagement. Instances of 'slow technology' (Hallnäs & Redström, 2001), such as those used in Lightphone and Cold Turkey, modify user experiences to bolster user self-control. Reflective and critical design methodologies can also be used to craft intentionally “unfriendly” user experiences (Dunne, 2008) that curb certain interactions to stimulate reflection, as demonstrated by Khovanskaya et al.'s unsettling personal web data interface (2013). Further still, applications such as Chatroulette combined technology’s ability to seamlessly connect individuals with the social friction inherent in interacting with someone new.
The current literature lacks a cohesive model for explaining different kinds and purposes of positive friction in user experience design or how recent advances in AI may harness or benefit from being seen through the lens of positive friction. This paper aims to understand positive friction from a behavioral perspective in the form of a model that describes its various forms and explore how the paradigm of positive friction can guide and enrich the future design for AI user experience.
In the paper, we propose a behavioral model of positive friction which is composed of two intersecting axes that create four quadrants, each of which describes a common form of friction that is beneficial in user experience (Figure 1).
This first dimension presents a tension between dampening and stimulating actions, in which the former imposes friction to lessen impulsive behaviors or make them harder to satisfy, and the latter employs friction more generatively to activate new ideas or cultivate happy accidents.
The second dimension denotes whether the purpose of friction is to prompt specific behaviors or, conversely, to upend business-as-usual mental models. Where the former’s purpose is typically to get behavior back on track by reasserting intentionality, the latter case the goal of design is to expand, rather than narrow, possibilities.
This results in four types of behavioral strategies and opportunities for UX application that address both users’ expressed needs and their actual behavioral tendencies, as well as the increased potential for communal social good:
Saving me from myself: Increasing self-control to achieve goals
Questioning assumptions: Disrupting auto-pilot behaviors
Stimulating action: Prompting or motivating movement
Deprioritizing efficiency: Embracing exploration and divergence
AI’s computational capabilities and common use cases have historically emphasized its ability to reduce traditional forms of friction for user benefit. However, its other attributes—the potential for surveillance, algorithms’ tendencies to reflect or amplify bias, and ethical questions about when to use AI—suggest that adopting AI and machine learning technologies without sufficient awareness of how they operate can also lead to harm. This means that even while AI displays impressive abilities to reduce friction within user experience design, in practice it may actually benefit from greater kinds and degrees of friction to ensure its ethical and unbiased use. As a result, we propose that this positive friction model may be usefully employed not only to describe friction in the context of user behaviors, but also as a way to understand how adding friction plays a critical role in designing and building AI interfaces, products, and services.
Comments