The Friction We Need
Most discussion about AI's impact on human capability focuses on jobs: which roles will be automated, which augmented, which created. A paper published in Communications Psychology by psychologists from the University of Toronto takes a different and more unsettling angle. The authors — Emily Zohar, Paul Bloom, and Michael Inzlicht — argue that the most significant long-term cost of AI systems making tasks too easy may not be economic but psychological: the erosion of the effort, struggle, and friction that make learning deep, creativity genuine, and relationships meaningful.
The paper, titled "Against Frictionless AI," does not argue that AI tools are useless or that convenience is harmful. It argues for a distinction between productive friction — manageable difficulty that drives growth — and unproductive friction — obstacles that add burden without benefit. Its concern is that AI systems, in their current design trajectory, are removing the former along with the latter.
Desirable Difficulties
The psychological research underlying the paper's core argument is well-established. Cognitive scientists have documented for decades that effortful learning — working through problems, encountering obstacles, generating explanations — produces better long-term retention and more flexible understanding than passive absorption of presented information. This principle, known as "desirable difficulties," runs directly counter to the design philosophy of AI systems that aim to deliver answers as quickly and completely as possible.
"We define friction as any difficulty encountered during goal pursuit," Zohar explained in an interview. "In the context of work, it involves mental effort — rumination and persistence, staying on a problem for some time, and this helps solidify the idea and the creative process." AI systems that complete the whole task from a single prompt — bypassing the intermediate steps where learning and consolidation happen — produce better immediate work products at the potential cost of the cognitive development that struggling would have generated.
The Social Dimension
The paper's more provocative claims concern interpersonal relationships. The authors argue that human relationships involve inherent friction — disagreement, compromise, misunderstanding, and the experience of not always getting what you want from another person. These experiences, they argue, are not bugs in human social life but features: they teach perspective-taking, build tolerance for other viewpoints, and train the social capacities that make deep relationships possible.
AI systems designed to be responsive, agreeable, and never frustrating create a fundamentally different social experience. "If you're used to an AI reinforcing all your ideas and being sycophantic, you'll come into the real world and you won't be used to seeing other ideas," Zohar says. "You won't know how to interact socially because you'll expect people to always be on your side." The concern is most acute for adolescents, who are at a developmental stage where navigating genuine human complexity is most formative. Young people who outsource significant portions of their social and cognitive development to AI may emerge with genuine deficits that no amount of AI-generated social coaching can later address.
What Makes AI Different From Past Labor-Saving Technology
A common objection to this argument is that new technologies have always removed effort — calculators from arithmetic, washing machines from laundry. The paper acknowledges this history but draws a key distinction: previous labor-saving technologies primarily removed physical or mechanical effort from tasks where difficulty was not the point of the activity.
AI is different because it is increasingly removing effort from activities where the difficulty is not incidental but integral. Writing involves struggle precisely because working out what you think and how to express it are inseparable activities — the struggle to find the right words is the process by which ideas are clarified and tested. Outsourcing that process to an AI produces a better output but bypasses the mental work that writing was doing for the writer's understanding.
Toward Friction-Aware AI Design
The paper does not argue for removing AI tools from education or professional contexts. It argues for a design philosophy that preserves productive friction rather than optimizing it away. "Instead of just jumping to the answer, it's more of a process model where it helps you think about the problem and teaches you along the way, so it's more collaborative rather than a one-stop shop for the answer," Zohar suggests.
Such a design philosophy would require AI developers to think about the long-term cognitive and social effects of their systems' defaults — not just the immediate user satisfaction metrics that typically drive product decisions. Whether market forces will create incentives for friction-preserving AI design, or whether the competitive pressure to deliver frictionless experiences will continue to dominate, remains an open question with significant implications for how a generation of AI-native users develops their cognitive and social capabilities.
This article is based on reporting by IEEE Spectrum. Read the original article.
