Appendix: Enshitification
We typically expect technology to evolve towards a better user experience; indeed, this has largely been the case throughout the development history of most digital technology. For decades, most developers and technologists broadly understood that their role was to make things better and, by doing so, improve the lives of ordinary people through exposure to superior tools and services. They held a firm belief that technology was fundamentally an enabler, and that contributing to its advancement was a way of contributing to society at large. Google’s famous unofficial motto, “Don’t be evil,” had been part of its corporate code of conduct since 20001, and it captured something genuine about how the industry saw itself; technology companies were not merely building products, they were building a better world.
This position has persisted even as Silicon Valley became a highly monetised ecosystem driven by shareholder returns and venture capital. Technology workers have consistently expected to make contributions in a broadly progressive way, and when asked to build products they considered harmful, many have refused outright.
In 2018, many engineers at Google feared that the technology being developed under Project Maven would be used to single out targets for killing, and a growing number of workers concluded that the programme was not in keeping with Google’s values.2 Google employees succeeded in getting the company to agree not to renew its deal to help the Pentagon build artificial intelligence tools for drone warfare.3 That same year, Google faced a second revolt when details emerged of Project Dragonfly, a censored search engine designed for the Chinese market. Employees joined Amnesty International in calling on Google to cancel the project, stating: “Our opposition to Dragonfly is not about China: we object to technologies that aid the powerful in oppressing the vulnerable, wherever they may be.”4 About 1,400 Googlers signed an internal petition, and two employees working on Dragonfly were so disturbed by the secrecy that they quit the team over it.5
At Amazon, the objection was similarly targeted at a specific product. Employees were up in arms about the company’s decision to sell its Rekognition facial recognition software to police departments and government agencies.6 In an internal letter to CEO Jeff Bezos, they were unequivocal: “Focusing solely on shareholder value is a race to the bottom and one that we will not participate in. We refuse to build the platform that powers ICE, and we refuse to contribute to tools that violate human rights.”7
At Microsoft, employees published an open letter declaring: “We are alarmed that Microsoft is working to provide weapons technology to the US Military, helping one country’s government ‘increase lethality’ using tools we built.”8 The letter called on Microsoft to cease working on “any and all weapons technologies,” to create a public acceptable use policy, and to establish an independent, external ethics review board.9 Microsoft CEO Satya Nadella pushed back on the protest, saying he was open to discussing corporate social responsibility but that the company would not withhold technology from democratic governments.10
The concern extended beyond military applications to the products people use every day. Former Apple, Google, and Facebook employees formed a group to take a stand against addictive tech; one of the group’s leaders, former Google design ethicist Tristan Harris, said he had been privy to “deliberate decisions” made by tech companies “that do great harm.”11 Harris’s internal presentation at Google struck a chord, spreading to some 5,000 Google employees, including senior executives.12 These were not outside critics; they were the very designers and engineers who had built the features in question, from the pull-to-refresh mechanism to the notification systems that keep users compulsively returning. Computer scientist Ben Shneiderman described this as “a historic moment of the employees rising up in a principled way, an ethical way and saying, we won’t do this.”13
In a survey for Wunderman Thompson Intelligence’s “Generation Z: Building a Better Normal” report, 74% of young people said they refuse to work for a company that goes against their values.14 The sentiment clearly runs far deeper than any single company or product.
Perhaps symbolically, it was also in 2018 that Google quietly removed its “Don’t be evil” motto from the preface of its code of conduct; the new document featured just one reference to the old motto, buried in the very final paragraph.15 The timing was not lost on observers. Some employees had suggested that Project Maven went against the “Don’t be evil” motto, but not being evil clearly was no longer a stated focus for the company.
The New Corporate Reality
And yet the landscape for such activism has grown more hostile; some of the tech industry’s biggest corporations are cracking down on employees who criticise their policies, stifling worker activism that was once ubiquitous in Silicon Valley.16 Mass layoffs, once rare for Silicon Valley giants, have become a routine part of workforce management, and a tougher stance on whistleblowers combined with a poor job market in tech is discouraging employee activism. The ability for corporations to push back against their own workforce is due in no small part to the advent of AI, along with a management misunderstanding that AI can simply replace people. This view, while still prevalent among tech entrepreneurs, is proving increasingly incorrect as AI begins to produce poor technical outcomes; combined with the over-recruitment of technical staff during the post-COVID boom, it has nevertheless given management a cudgel with which to silence dissent.
This new corporate reality has begun to mean that we are seeing a regression in the quality and improvements we would typically expect in digital products. Making a user’s life worse without them realising it is the new corporate reality, and this is typically being done so that more advertising and engagement can occur. This is the crux of Cory Doctorow’s argument for “enshittification.”
Enshittification
The term was coined by Doctorow in November 2022, and the American Dialect Society selected it as its 2023 Word of the Year.17 Doctorow has also used the term “platform decay” to describe the same concept.18 As Doctorow himself puts it: “Enshittification is a theory that explains the worsening of the things around us, upon which we depend.”19 More precisely, it describes a three-stage process by which digital platforms systematically degrade:
“First, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die.”20
For enshittification to occur, there must be at least two parties and a platform sitting between them.21 These platforms are increasingly technology companies, and they are usually selling advertising or enabling others to sell their own products via their infrastructure.
Google Search
As an example, consider what has happened to Google Search. In February 2019, Google’s search team faced a crisis: a leaked internal “code yellow” alert, later cited in the DOJ’s antitrust lawsuit, revealed declining revenue from search ads.22 Ben Gomes, a search veteran who had been at Google since before they even had ads in search results, resisted the ads team’s push to prioritise growth metrics over search quality, arguing that “growth is all that Google was thinking about” and that they shouldn’t compromise user experience just to boost engagement.23 Gomes left his position as head of search about a year later, and was replaced by Prabhakar Raghavan, who had previously served as Google’s head of advertising.24 Industry analysts claimed that Raghavan’s background in advertising, rather than search technology, had led to decisions prioritising revenue over search quality.25
The result is that search results have been made suboptimal; users now frequently need two or three searches to find what they are looking for, ensuring that Google can sell advertising multiple times over. While this de-evolution is in some respects not immediately obvious to users, there is a growing sense that Google is simply not as good at search as it once was. “It’s the worst quality results on Google I’ve seen in my 14-year career,” says Lily Ray, senior director of search engine optimisation at digital marketing agency Amsive Digital.26 A 2023 survey by Pew Research found that 62% of frequent searchers believe search results have worsened over the past three years.27 But Google holds a de facto monopoly on search, and so most people do not switch. Google has consistently been getting worse in both user experience and search results for years now, but the frustrating thing is that Google has been, and still is, so incredibly dominant that there really isn’t any viable competition.28
Amazon likewise functions to an extent as a search engine; it is essentially a local search interface for a database of products and sellers. However, to reach the top of the listings, sellers must pay more for the service and guarantee that their products will not be cheaper elsewhere. In this case, both the user and the seller are disadvantaged: the user receives incorrect search results based not on relevance but on a seller’s ability to pay, and the seller is penalised in that they must pay simply for the privilege of being found. This is compounded by the fact that Amazon, like Google, holds a captured market share and is the de facto online retailer for a great many people. In Doctorow’s original post, he discussed the practices of Amazon; first, Amazon started selling goods below cost to build up a user base29, and then, once that base was locked in, the platform began extracting value from both sides of the marketplace.
Likewise, Facebook as a social media giant offers few alternatives; users simply put up with the advertising because there is nowhere else to go. And enshittification also predicts that as well as the user having a worse experience, so too, once “captured,” do the advertisers, who now pay more for less and with diminishing control over the statistics coming back from their unique impressions. This is the highest form of enshittification: when platforms abuse their business customers, and a lack of meaningful competition and regulation means they can still increase profits.30
Enshittification and UX
So how does this fit in with UX? Simply put, user experience presumes and strives to achieve a good experience for the user, with the understanding that if the user has a bad experience they are likely not to use the product further. Enshittification only works when you are the platform and you have captured enough users on both sides of the market (as both buyers and sellers, or as both audiences and advertisers) so as to make using the platform a necessity and moving away from it difficult. Consider how software development itself can be affected by this dynamic: both through the various app stores, which are captured markets in their own right, and through the developer tools we may wish to use that require subscriptions and operate on an online platform that initially makes your job easier but will eventually demand payment for that convenience.
For UX specifically, enshittification as Doctorow defines it may not be the most precise description of what we observe; possibly regression or deterioration of the user experience is more fitting, in that from our perspective we are focused on the end user’s experience of the interface and the interactions surrounding it. My own definition of User Experience, which I have held since 2022, is that it:
“is an umbrella term used to describe all the factors that contribute to the quality of experience a person has when interacting with a specific technical artefact, or system. It focuses on the practice of requirements gathering and specification, design, creation, and testing, integrating best practice, heuristics, and ‘prior-art’ whereby the outcomes can be qualitatively evaluated using small numbers of users, placing humans firmly in the loop.”
But there is an implied understanding within this definition that the objective is for users to rate their experiences highly when comparing an evolving UX; that if we could detect signs of deterioration or regression, the new interface or interactions would not be implemented, and so we could expect improvement or, at worst, no change. Enshittification goes directly against this implied progress, in that after a UX cycle the objective would deliberately be for a reduced user experience as leverage for increased platform profits. UX as a discipline is directly opposed to this way of thinking, grounded as it is in the tools, statistical analyses, and methodologies that practitioners employ, and in the mindset of the specialists themselves, for whom a better user experience is always the goal. One cannot credibly claim to be a UX specialist while accepting a worse interface as a deliberate outcome; this is especially true given that so many of the methodologies we employ (heuristic evaluations, usability testing, accessibility audits, and the like) exist precisely as substitutes for the user’s own voice, standing in where the user cannot advocate for themselves. To use those methods in service of degrading the very experience they are designed to protect would be a fundamental betrayal of the discipline’s purpose.