A Runaway Experiment

By Robin Mansell and Ed Steinmueller

The commercialisation of the internet in 1995 was widely greeted with enthusiasm about its potential for improving access to information, fostering new business opportunities, and providing a forum for political debate. Twenty five years later, the internet is linked to ‘fake news’, is dominated in many ways by large platform companies and is implicated in the polarisation of political debate. What went wrong?

In moving from a publicly supported information infrastructure for research to a commercial marketplace, early decisions created a great opportunity and few restrictions for commercial development. These decisions were unprecedented. Earlier forms of information provision — newspapers, magazines, radio and television, were all regulated. For example, the precursor to the US Federal Communication Commission, the Federal Radio Commission established the standard that a radio station could be licensed only if it could be shown to be ‘in the public interest, convenience, or necessity’. With the internet, a long legal history of newspaper and magazine regulation through libel and defamation laws was bypassed by devolving liability from the information provider to the author. Public authorities also moved quickly to provide standards for online payments and to ensure that contracts were legally enforceable, but largely unsupervised. Collectively, these decisions amounted to a grand experiment in ‘free market’ economics. This experiment was not about a market for t-shirts, candy bars or toothpaste. It was an experiment in what would become the dominant means of interpersonal communication and information publication.

A principal outcome of this experiment was to enable a digital platform business model, a means of earning money, based upon increasingly precise observation of user behaviour that could be used to improve the reach and targeting of advertisements. All the large platform companies aside from Amazon have been built on advertising revenue. For the most part, this advertising revenue is not new money, but money diverted from other advertising channels – newspapers, magazines, television and radio – with the effect that these traditional media channels have fewer resources for engaging in journalism and other content production. Ironically, what some platforms have taken away, others are restoring. Netflix and Amazon are investing in content production based upon subscription business models. This is what we might expect of a successful experiment in the operation of a ‘free market’; competition displaces some and rewards others, a consequence of user choice, but this is not the whole story.

Experiments often come with surprises. The rise of the giant platforms was such a surprise. While it was recognised that network effects (joining a larger network offers greater opportunity for connection) would be important, few anticipated that these effects would continue undiminished as the platforms attracted millions and then billions of users. This was possible because of another surprise. After a long hiatus, practical applications of AI made it possible to automate the selection and targeting of advertising and content as well as the surveillance and appraisal of users. These applications greatly reduced what traditionally had been a source of diversity in media, differences in audiences.

With AI, it became possible for a single provider to provide content for diverse audiences. This opened the door not only to growth, but also to some of the ills we are now experiencing. An AI algorithm tasked with capturing audience attention is indifferent (unknowing) of any concept of value other than advertising value. Televised executions, conspiracy theories, and Presidential tweets generate audience attention and advertising value.

We might conclude that this offers an agenda for reform. Might codes of practice or revisions in the AI algorithms set ‘filters’ that would mitigate the problems? In principle, there is some scope for this. However, the experiment largely has outrun capabilities for human intervention. In one minute, YouTube receives 500 hours of content, Facebook receives 317,000 posts (status updates), and 1.7 quadrillion (10^15) bytes of data are added to the internet. Filtering this amount of data is a daunting prospect and it would introduce numerous unintended effects that would constrain free expression.

The internet is a runaway experiment with both positive and negative consequences. These are most directly observable in the behaviour and performance of the large digital platforms. In our book Advanced Introduction to Platform Economics we examine how platforms have emerged, what the options and limitations are for self-regulation, and the policies that might re-set the experiment to achieve public values of public interest, convenience and necessity. Among the most urgently needed policies are those that directly tackle the economic incentives created by the current business models which nudge people towards attitudes and behaviours that are either illegal or harmful to themselves or others. The answer lies not in the top down (and disputed) censorship of content with regulation constantly running behind online developments, but in a fundamental restructuring of who owns data, who controls it, how money is generated from it and how those who do so are held to account.

Advanced Introduction to Platform Economics

Robin Mansell, London School of Economics and Political Science and W. Edward Steinmueller, University of Sussex, UK 

Request a free ebook inspection copy here.