Here’s an easy prediction: 2018 will bring mounting pressure to develop large-scale, automated responses to online misinformation. And here’s a hopeful one: That pressure will spark an increasingly frank discussion among journalists, policymakers, and platform companies about how to bolster the fact-building institutions that anchor public truth claims.

The quest to automate online fact-checking began well before the furor over so-called “fake news.” Fact-checks have a consistent structure built around discrete data elements — a claim, a claimant, a verdict — that lends itself to marrying human and machine intelligence in interesting ways. One of the first attempts to do this was Truth Goggles, an MIT project that (as initially conceived in 2011) tried to harness the work of professional fact-checkers to build a “magical button” that would instantly flag false claims on any web page.

Since then, fact-checkers and computer scientists have worked together on a string of projects that aim to automate different part of the fact-checking process. One thing these efforts have in common is using automation as an enhancement, rather than a replacement, for journalistic work — as what Bill Adair, writing in these pages past year, called a “force multiplier” to help human fact-checkers keep up with a growing tide of online misinformation.

For example, the most tedious part of a fact-checker’s job is hunting for interesting and important claims to check; a lot of what politicians say in speeches and debates turns out to be vague rhetoric or statements of opinion. ClaimBuster, developed by computer scientists at the University of Texas at Arlington, pores over transcripts to identify and rank factual statements that fact-checkers might want to investigate. (It also retweets checkable claims here.)

Another promising tool, being developed by the U.K.-based Full Fact, scans media feeds to track which politicians and news outlets are repeating claims that have already been debunked. Called Trends, the project is designed to give the fact-checkers a strategic view of the misinformation landscape, so they can target their corrections more effectively. Political journalists will also be able to use it to guide their reporting.

The most powerful automation technology deployed so far is in some ways the simplest: Since 2016, growing numbers of fact-checkers around the world have been using a new tagging scheme, called ClaimReview, that makes their work legible to algorithms. This means that when Google recognizes the claim being searched for — try “the Russia investigation is a made-up story” — it can preview the verdict in a “snippet” at the top of the search results. (The same tagging system lets the Amazon Echo look up answers to factual queries.)

The research involved in efforts like these has started to bring into view the “holy grail” of an end-to-end fact-checking engine that can check at least some kinds of claims in real time. ClaimBuster and Full Fact have trials in the works (see videos here and here) which operate by matching new claims against fact-checkers’ databases — or, in certain cases, by consulting original data sources such as economic statistics or voting records. As Full Fact noted in a 2016 report, the key to these efforts “is to make sure the kinds of resources fact-checkers can rely on are available as structured data that computers can use.”

And that’s what’s most exciting about these initiatives: Designing the systems to automate verification will draw focus back to the public, institutionally sanctioned data sources that human fact-checkers depend on to certify facts about everything from hurricane strength to inflation rates. These include agencies at every level of government — the U.K.’s Office of National Statistics is an ideal-typical example — as well as countless scientific and civil institutions that set standards and produce benchmark data in different areas of research or policy.

Political fact-checkers couldn’t do what they do without these kinds of institutional resources. Of course, neither could Google’s search algorithms, which depend both directly and indirectly on institutional authority reflected in our patterns of clicking and linking online. In fact, professional fact-checking groups act as a kind of social engine — one of many — for translating institutional knowledge into data that computers can work with.

Reflections on the rise of “algorithmic authority” have tended to oppose it to the institutional kind, based on old-world mechanisms like formal credentialing and peer review. The push for “automated” fact-checking may help us to come to terms with a reality that has always been messier than that, making explicit the institutional subsidy that algorithmic intelligence depends on.