返回资讯中心
外部精选
软件工程
#开发者实时流

Friction: What have we been reducing?

An exploration of the role Friction plays in human learning, judgment, and the formation of perception. It examines how personalization and generative AI reduce Friction, and what we may be losing in the process.

Hacker News 最新juun_roh2026/10/18 · 2026/10/1830 分钟阅读

以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。

An exploration of the role Friction plays in human learning, judgment, and the formation of perception. It examines how personalization and generative AI reduce Friction, and what we may be losing in the process.

Historically, technology has advanced by lowering the cost of reaching our goals.

From carriages to trains, cars, and airplanes, the evolution of transportation has reduced the cost of travel. The cost of finding information has fallen as well, from library catalogs to search engines. Today, recommendation algorithms lower the cost of finding and choosing things, and generative AI goes a step further by lowering the cost of producing them.

The history of technological progress has, in many ways, been a story of reducing the costs we once accepted as necessary to achieve the outcomes we wanted. Time, effort, and inconvenience gradually came to be seen as unnecessary obstacles along the way. In the technology industry, these costs are commonly referred to as Friction, and technology has evolved to reduce or eliminate them. Indeed, much of that Friction was unnecessary. Nobody longs for the days when people had to walk for days just to deliver a simple message or copy an entire book by hand. Good technology made tasks easier; great technology made them feel effortless. In tech, "frictionless" naturally became a term of praise.

However, not all of the time and effort technology has eliminated is remembered as something we're glad to have been rid of.

Anyone who lived before mobile phones used to wait for someone who was running late, with no way to reach them. People remember the days spent waiting for a reply to a letter, or for film to come back from the lab. From the standpoint of efficiency, these were obvious inconveniences. Yet they are often recalled with fondness or a sense of fulfillment rather than annoyance, as if the waiting itself had made those encounters, letters, and photographs more meaningful.

This left me with a small but fundamental question. We have been asking how to eliminate Friction so that we can get the results we want more easily. But what, exactly, was it in the first place?

In 1994, psychologist Robert Bjork introduced the concept of Desirable Difficulty. Certain conditions that make learning more difficult can impair immediate performance while improving long-term retention and transfer. Spacing out study sessions rather than cramming, interleaving different types of problems rather than practicing one type at a time, and trying to retrieve information from memory rather than simply rereading it—all of these can feel harder and slower to the learner.

A related phenomenon had been observed earlier under the name of the Generation Effect. In a 1978 experiment, Slamecka and Graf found that people remembered words they had generated themselves from cues better than words they had simply read. The effort of generating an answer, rather than merely receiving it, helped make it more memorable.

The implication is straightforward: difficulty is not always a cost to be eliminated in learning. Something happens as we struggle to recall, become confused, make mistakes, and revise our understanding. Some kinds of difficulties are not merely obstacles to learning, but part of what makes learning possible at all.

When faced with an emergency, most of us hesitate to act. Various factors, including the bystander effect, can hold us back, from uncertainty about the situation itself to concerns about the responsibility we might bear if we intervene. Firefighters, however, are different. They assess situations and take action quickly, even in unexpected or dangerous circumstances. It is no exaggeration to say that a large part of what makes firefighters seem heroic is their ability to act without hesitation even when faced with the unexpected.

Personal characteristics and a strong sense of duty may certainly contribute to how firefighters respond, but they alone cannot fully explain the difference. Gary Klein studied how experts, including firefighters, make decisions and described this phenomenon through a framework known as Recognition-Primed Decision (RPD). Rather than systematically comparing all theoretically possible options to calculate the best course of action, experts quickly recognize familiar patterns in a situation and identify a course of action that fits. In other words, what appears to be an unfamiliar and uncertain situation to an ordinary person may be recognized by an expert as a familiar pattern, one they have learned to identify through experience.

Here, the key point is that the difference between experts and non-experts is not simply a matter of how much they know. Firefighters not only learn the theories and train to respond to various situations, but also live through real emergencies alongside their fellow firefighters. By observing how their colleagues respond and acting together in the field, they repeatedly experience success and failure firsthand, making mistakes and sometimes coming to regret their actions. These lived experiences shape the perception of the situation itself. What they notice, which possibilities they consider, and which actions they prioritize all change.

This shift in perception is not unique to experts. Research on attention and perception has shown that previous selection experiences can influence what we pay attention to and what we choose later on. In other words, past experiences leave their mark on future choices.

This theoretical dichotomy, however, fails to explain a growing number of cases in which neither current goals nor physical salience can account for strong selection biases. - Top-down versus bottom-up attentional control: a failed theoretical dichotomy

In cognitive psychology and neuroscience, attentional control has traditionally been understood in terms of two broad mechanisms: top-down and bottom-up control. Top-down control refers to the voluntary allocation of attention based on current selection goals, while bottom-up control refers to attention being drawn by salient external stimuli.

However, in their paper Top-down versus bottom-up attentional control: a failed theoretical dichotomy, Awh et al. argued that strong selection biases exist that cannot be explained by this theoretical dichotomy. They identified two factors that challenge this dichotomy:

Awh et al. explained attentional selection in terms of a priority map shaped by current goals, selection history, and stimulus salience. Although this research focuses on visual attention, we can reasonably infer that our past choices and their outcomes influence the choices we make in the future.

In a 2026 study, Recognize danger, but disregard it: a mixed-methods analysis of the knowledge-action dissociation in volunteer firefighter hazard recognition, Baron et al. introduced the concept of Rescue Impulse to describe a tendency observed among volunteer firefighters. When firefighters encountered a person in danger, they often moved to intervene automatically, before consciously evaluating the hazards or considering their own safety. This tendency to initiate rescue actions before conscious deliberation may offer another example of how patterns of selection shaped by past experience can influence perception and behavior before deliberate judgment takes place.

What we develop through learning might be more than just a body of knowledge. I would say that one of the outcomes of learning is the formation of a Selection Structure that shapes what we notice and what we choose—a structure that operates beyond conscious decision-making, influencing our attention, perception, and even behavior.

We can see this in sports as well. Experienced athletes often struggle to teach beginners how to move the way they do. These movements have become so familiar through years of practice that athletes perform them almost without thinking, and may no longer know how to break them down into distinct steps. What they develop through practice is not simply an understanding of how to perform a movement, but an intuitive framework for recognizing what matters in a given situation, responding to relevant sensory cues, and adjusting their movements accordingly.

If this is the case, learning may ultimately produce structures that are difficult to articulate and cannot be fully conveyed to others through declarative knowledge alone. The Friction we have been trying to eliminate may have been part of how these structures took shape, in ways we never noticed.

Recording technology has enabled us to listen to music whenever and wherever we want. The development of transportation has tremendously reduced travel time. The invention of writing systems and the development of printing technology have greatly contributed to the spread of knowledge. However, the Friction that these technologies were supposed to eliminate has not completely disappeared. People still attend live concerts, ride bicycles or take trains instead of planes, and even choose to travel on foot. We still attend classes even though books are available.

Although technology has undoubtedly reduced Friction in our lives, some inconveniences that were once considered obstacles to efficiency have persisted, acquiring value in their own right. Even when more efficient options are in our hands, we still choose to embrace inefficiency. Earlier technologies had a fairly clear target: they made specific things more efficient, and traces of the Friction they reduced could still be seen in other forms.

However, which kinds of Friction personalization removes, and what that changes, remain opaque.

As algorithms have become quietly embedded in our daily lives, personalized content recommendations based on users' activity records have become the norm in the content industry. Algorithms, which rank and recommend content based on user data, have let platforms implement content prioritization and optimization policies in real-world systems. The tech industry came to treat even the act of scrolling past without selecting anything as Friction to be reduced.

Personalization does more than help us find what we want. It shapes what we encounter before we even begin to choose. Content that ranks low in an algorithm's recommendations may remain effectively invisible unless we actively seek it out.

In doing so, personalization optimizes not just our choices but the conditions under which we make them. It reduces the effort required to find what we already know we want, but it also limits our opportunities to encounter the unfamiliar—to develop new preferences, revise old ones, and build the selection histories that shape our future choices. Over time, the world we experience may come to reflect less of what exists than what platforms have decided is worth showing us. And we rarely notice which possibilities have quietly disappeared.

But reduced Friction may shape more than the choices that follow. AI researcher Connor Leahy points to an example in a recent interview: dating apps have done more than facilitate encounters; they may have reshaped the social norms governing how people approach strangers and how such approaches are perceived. He adds that people have increasingly come to see approaching strangers in real life, outside the context of dating apps, as awkward or inappropriate.

The phenomenon he describes can also be viewed from a slightly different perspective. As the forms of Friction we have traditionally encountered have diminished, perhaps we have also had fewer opportunities to learn how to navigate them. Someone used to GPS may feel burdened, even reluctant, when asked to find their way with a paper map. Have we, in the same way, come to see the Friction of dealing with other people as a heavier burden than before?

Fashion comes full circle. This expression captures the cyclical nature of fashion trends, commonly reflected in the revival of styles from the past. However,

正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。

打开原文
去社区讨论 →