A new truth layer for the web

Over the years this idea has attracted entrepreneurs and technologists, and so far no one has been able to figure out a workable, widely-adopted product.

The problem to solve is obvious: With so much content being published online, it’s difficult for most people to determine the quality and credibility of a given webpage or other piece of content. How can you know if the article you’re reading has incorrect facts, is incomplete, or was produced by an organization with serious ethical issues? Isn’t there some way to compare all the articles and content on a given topic and surface the best, most accurate and complete version?

A team of 16 people in Paris are the latest to try and solve this problem. Their product is Trooclick, and it will launch an initial version this month. (Today, at the Global Editors Network Summit in BarcelonaI’m moderating a panel about fact-checking that includes Trooclick CEO Stanislas Motte.)

Trooclick is a browser plugin (Chrome and Firefox) that alerts you if an article you’re reading includes what they call “glitches.” A glitch could be an incorrect fact, information that conflicts with other media reports about the same topic, or something about the publisher’s ethics, or the ethics of the article itself, that a reader should be aware of.

“These are warning signs that something in the article doesn’t quite match with a public database, or with other articles that have been written about that same subject,” said Robyn Bligh, a translator with the company who also leads its communications efforts, in a phone interview. “We’re not saying that it’s completely false; it’s a warning sign.”

During a demo they showed a Venture Beat article about a company’s IPO. Here’s a look at the Trooclick window that popped up to tell me about the glitches:

Trooclick flagged it due to the fact that a Wall Street Journal article included a different amount of money that a company was planning to raise in its IPO. Note: the “Is this article reliable” feature is a user-generated voting aspect that wasn’t active when I used People will be able to vote an article up or down.

The app noted that the company’s IPO filing with the Securities and Exchange Commission contained a different amount for the IPO raise, as a well as a different filing date:

Their strategy is to start by focusing on identifying glitches about financial/business news.

“The main target is professional in the financial and economic fields because they’re the ones who can benefit the most,” Bligh said.

The thinking is that if Trooclick can help a business or financial professional ensure they always see correct information in news articles, then the company sees that as a path to gaining a foothold among users. They also see potential in the future to have others use their technology to surface the best content, and pay a licensing fee to do so.

“The ambition is to check all the field of the news,” said Pierre-Alber Ruquier, a former journalist who is the company’s CMO, in the interview. “So for the moment we do it step by step. It’s more a question of priority of which [subject] to start with.”

Business and financial news is important for many professionals, so that’s where they will start.

How it works

The way Trooclick works is relatively simple to explain but harder to execute: it analyzes the text of the webpage you’re on and compares it to their database of facts and figures to see if anything is related between the two. If there is a match, they see if the data they’ve collected is different from what you’re reading. If that’s the case, it alerts you to the glitches.

As for the ethical glitches, they will have a set of things to look for, such as the use of anonymous sources. For example, a blog post by Trooclick notes that a recent article in TechCrunch would have likely included some “media ethics conflicts” notifications given the number of anonymous sources:

Reading Techcrunch’s article, Trooclick was surprised to see the number of times they mention anonymous sources. The article is full of “we’re hearing from multiple sources”, “we hear that”, “people said”, will apparently”, “a source tells us” and so on. These expressions of uncertainty are among the criteria which Trooclick will be able to analyze in the future when rating there liability of a news article.

By the end of this month, they expect their system to capture and build a database of roughly 30 different economic properties that will be used to compare against an article you’re reading.

This data will be drawn from SEC filings, official company press releases, and other data sources they deem reliable. They also extract key data from news articles published by a growing list of close to 100 publishers whose content they scan on a constant basis. (Trooclick doesn’t store the articles themselves; it extracts the key data from the article, such as share price, and stores that in a database.)

The challenge of a truth layer

As noted above, there have been several attempts to figure out the right truth/quality layer for the web.

Two projects that didn’t fulfill their initial promise were NewsTrust, which was meant as a way for people to collaboratively rate the quality of news articles, and to come up with the best coverage on specific topics. It ended in 2012 and the product and company domain were transferred to Poynter.

Another attempt in the same vein was NewsCred. It initially launched as a project to surface the best news articles using a mix of technology and user feedback. The company made a major pivot away from that and is today a leader in the content marketing technology space.

In terms of ongoing projects, Truth Goggles will use the PolitiFact database of fact checks, among others, to compare against a given article you’re reading. Like Trooclick, its consumer implementation would be as a browser plugin, but it hasn’t yet launched. There’s also LazyTruth, an effort led by MIT graduate Matt Stempeck to help identify urban legends and scams in your email inbox. It’s currently a Chrome extension, or you can use it by forwarding a suspect email to ask@lazytruth.com to get an analysis.

Another effort is Skeptive,which relies on users to identify conflicting claims online. Skeptive then attempts to determine “which side of a dispute is most supported by the sources that any given User trusts. Put together, these two processes tell you whether there’s a source out there that you trust that disagrees with the sentence that you’re reading.” Their goal is to find a better way to resolve online disputes and differences of opinion.

Trooclick and its ilk typically have two big challenges:

  • Building out a big enough database of quality data to compare against what people see and read.
  • Getting enough people to install their plugin/use their app.

The two elements are obviously connected. You need the data to deliver a good user experience in order to drive adoption. But without adoption, it tends to be hard to raise money to keep building out the data and product.

Trooclick is attacking the scale/adoption issue by focusing on a one area where there is potentially real value to users, and where they can access or build databases of relevant facts. Based on the alpha, their technology works and is nicely implemented for the user.

One challenge for them, and anyone else who relies on a browser plugin, is the fact that more and more reading is done on mobile phones, which renders most browser plugins useless.

Hey, nobody said building a truth layer was easy. That’s why people keep trying.

We have made it easy to comment on posts, however we require civility and encourage full names to that end (first initial, last name is OK). Please read our guidelines here before commenting.

  • http://www.lexalexander.net lexalexander

    Automating the process of fact-checking is going to be lengthy and iterative. But I see this as an encouraging step in the right direction, particularly with its shortcomings acknowledged up front, as here.