The Factual
Reasonable People #41: I haven't felt optimistic about news for a long time, but learning about The Factual kindled some hope
The Factual is a algorithm-driven search engine which is doing something interesting to realign the incentives around news and news reading. They have a podcast, trending topics feed and newsletter, but the heart of it is an algorithm which rates news stories for credibility.
Individual news articles are scored based on four criteria:
Quality and diversity of sources
Article tone
Author expertise
Site quality
This is not a fact-check. The score doesn’t tell if you if the story is factually correct or not, just the likelihood that it is reliable.
I learnt about The Factual from the Social Media and Politics Podcast, where co-founder and CEO Arjun Moorthy lucidly explained to host Michael Bossetta the reasoning behind these criteria (listen here: SMPP #141).
Quality and diversity of sources: the first criteria work like the reverse of Google’s PageRank algorithm. Google tracks how credible pages are by how many other pages link to it, The Factual look at how many outgoing sources are linked or quoted in the article. More sources, and more diversity sources, add to an article’s score. From the interview I understood that this also includes quotes of different people, and these are also weighed (so, presumably, “Professor X of Harvard” is rated higher than “Mr X of Springfield”).
Tone: is assessed by computer language models (NLP, as they call it). There are established models for detecting the emotional sentiment of text, but The Factual aren’t interested if a text is happy or sad or angry. Rather, they give a positive score to straightlaced text which doesn’t contain sensationalist, emotive or otherwise salacious language
Author expertise: articles are categorised by topic and a record is kept by the Factual of 50,000 individual writers. Their algorithm rewards authors who have written on that topic before - essentially, as founder Moorthy puts it, the algorithm rewards beat writers rather than generalists. As I understand it, the algorithm also allows authors to build up credibility over time by writing articles which get high scores on the other criteria.
Site quality: this reflects the history of that site of publishing highly rated articles.
There are a couple of interesting things about how The Factual does things.
The first is that they set themselves aside from other sites which republish links by NOT optimising for engagement. The facebooks and twitters of this world tune their algorithms to keep people online, clicking links and reacting to other’s comments. The Factual want people to click on links, but to hear them tell it they also want people to leave; a key part of their vision of their audience is readers who want to get briefed on the facts as quickly as possible without spin or delay. Moorthy puts it like this (I paraphrase from memory) : You free up an enormous space of design choices once you give up on solving all problems to maximise engagement.
Having opted out of the engagement game, The Factual can also opt out of the identity expression game. The Factual discussion section uses anonymous comments, but unlike many sites where anonymous commenting frees people to be the worse version of themselves, here it may allow people to leave behind some of the performative, tribal, attention seeking, clout chasing behaviour which makes social media so polarised. The trick is that they allow established users to give reputation to comments: users can mark that they “respect” a post, which is similar to a like or favourite on other sites but the connotation is merely that you respect what someone said, not that you like or agree with it. Importantly, although users see how many other people respected a comment, the total number of endorsements is not displayed. You can’t use your likes or followers as a status play on the forum. Rather, they have created a reputation system similar to the one used by stackoverflow. Those who are established are allowed greater power, including to give respect to new users, and in this way the whole system bootstraps credibility by supporting tracking of reputations across time.
The second interesting thing about The Factual is the limit they put on community judgement. As far as I can tell, and in contrast to sites like Reddit etc, they use their algorithm to rate article credibility, and support community discussion of articles they highlight, but they don’t feed user reactions back into their algorithm. This is not a crowdsourcing site. They don’t cede judgement about credibility to the mass of users, and so are free from tides of partisan or emotive passion. Crowdsourcing so often crystallises out of the mass of reasonable individuals the lowest common denominator of our basest instincts - a need for lazy titillation and cheap emotional shots.
Moothy, in the podcast, makes an analogy with automation in cars. Absolutely, he says, we want automation to help people drive more safely, but ultimately the driver has to know that they have the final say, and the final responsibility. I don’t know if he was aware, as he said this, that his site honours this thought from both directions: there is a strong algorithmic design, but The Factual seems to have a clear vision for both the task they are trying to do and the limits of that task. Articles are scored, but the site doesn’t pretend to offer definitive judgement, or even give you the single best, highest scoring, article on a topic. Instead it scrupulously offers multiple articles, from different perspectives, on a topic. But also, by not going down a pure crowdsourcing route, The Factual doesn’t crowd out individual judgement from the other direction, overwhelming individual discernment with raw social power.
I’ll be watching The Factual closely, and hoping for a future where individual journalists adapt their articles to get high The Factual scores - where they take out emotive language, and write longer pieces which quote a wider diversity of sources, in order to be rated highly by algorithms which serve our wider, wiser, selves rather than the shallow, poor impulse control, versions of our selves brought out by the engagement algorithms.
References
Website: thefactual.com
Social Media and Politics Podcast: #141: Rating News Credibility with Algorithms, with Arjun Moorthy
Also in reasons to be cheerful about journalism: Sheffield Tribune and the local news subscription model
I wrote about The Factual before I realised that there was another time I had felt optimistic about news journalism recently, but it had snuck up on me. The Tribune is “Sheffield’s new independent quality newspaper, delivered via email”. Like The Factual it uses a subscription model, and has used this to generate regular, high quality, local reporting. As a local I recognise that the writing will have limited appeal to people outside the city, but it is interesting and important The Tribune are making the model work, doing what looks like traditional journalism in a world where traditional newspapers and traditional journalism jobs have been cut and cut and cut - particularly in local news. And they are expanding - The Tribune was recently able to hire a second journalist, giving founder and editor Dan Hayes the ability to take a well earned week off without readers missing out on regular updates.
Yesterday The Tribune published this piece, which is a great example of what they can do, and perhaps of general interest to readers of this newsletter
We’ve probably all heard the stories about WhatsApp rumours leading to lynching in rural India. The stories suggest something dark about the power of social media to turbocharge gossip and rumour, but this darkness isn’t confined to places which might sound exotic and far away to some Western readers. A few miles from where I’m writing the same dynamics of rumour, charged by social media platforms and conditions of poverty and vulnerability, recently generated similarly violent outcomes.
The Tribune has the details.
In other news
NEWSLETTER: Stranger Apologies from Kevin Dorst
“A blog about why people are more rational than you think”
Kevin’s take on the danger of “irrationalist narrative” is very close to my own. So, naturally, as the authors of two newsletters on very similar themes about rationality and argument we are planning to devolve into narcissistic bickering and embarrassing mutual recriminations over the next few months, firing pot shots at each other from our respective publications in front of an increasingly bewildered and dwindling audience. Watch this space!
PODCAST: Behavioural Science Uncovered
Information-calorie dense discussion of behavioural science research.
I enjoyed this episode: Self Persuasion with Peter Schwardmann and Egon Tripodi which discusses this paper: Self-Persuasion: Evidence from Field Experiments at International Debating Competitions which really deserves a write-up of its own.tl;dr really thoughtful and impressive field experiments using debating competitions which shows causal evidence that arguing for something can change your mind in favour of that position (and importantly extends this phenomenon outside of lab settings).
And finally…
“A sentence is a device for taking a thought for a walk”
Cartoon from Martin Shovel.
Twitter: @MartinShovel
Web: shovel.co.uk, creativityworks.net
Comments? Feedback? Respect? I am tom@idiolect.org.uk and on Mastodon at @tomstafford@mastodon.online