Data, data, data. From black and white to colour.
Overtone looks at Facebook, recommendations, and privacy
Hello, this is Christopher Brennan, and this week’s edition is a “Meta” one about algorithms, recommendations, data, and all the problems they can cause. The most prominent example in recent months has been revelations about Facebook from whistleblower Frances Haugen, which include how the company ignored or sidelined information about the harms that its platform causes, from teen mental health to political anger.
Facebook’s focus on optimising for engagement reminds me of an analogy that my co-founder Philip and I use for how the internet works right now.
We sometimes say that looking for news on the internet is like a big market where everyone is buying apples (the fruit not the phone) from different sellers. The people buying know about the sellers, but don’t have that much information about the apples themselves, even whether they are Red Delicious or Granny Smith. It's as though everything is in black and white, on a grayscale.
One of the ways the sellers try to get you to take their apples, then, is by offering you the darkest apples possible, because they appear to be the tastiest.
This is similar to offering someone a news article or a post on social media because it is likely to get them to click. The places where people get their apples keep picking dark ones from their suppliers on the farms, and then offer them up to you and people like you based on the apples that you have chosen before, learning everything about you to hone in on the exact shade of dark for you.
Unfortunately, the dark apples that they serve all have different tastes and don’t always satisfy your needs. Worse than that, there are a bunch of “bad apples” out there, from lazy farmers or even purposeful bad actors who want to make people sick. But all consumers know about the apples themselves is where they are on the gray scale from white to black. At any given moment you can describe the taste of the apple you want, but can’t always find it.
What is necessary then, is a Wizard of Oz-style moment when all the fruit returns to Technicolor. If we know an apple is spotted or green or yellow then the people who eat them can make informed choices about what they want. Overtone’s scores are our first step at bringing back colour and telling us more about what an article *is* rather than how it's been interacted with. Lots more to come.
That analogy is a long way of describing what the problem is, though there is also a lot more to explore on the use of data, how it impacts people’s privacy and how it’s used online. This week’s list of recommended articles is about algorithms, and data, with recent pieces that were the apple of our algorithm’s eye.
Quality stories you might have missed that our algorithm suggests:
Facebook fails to curb the spread of hate speech in Ethiopia - The Mail & Guardian
John Edwards is off on a ‘late-life OE’ as UK privacy watchdog, and he’s not done with Facebook yet - The Spinoff NZ
TikTok’s Next Big Move? To Become Facebook - Wired
Lloyd v Google: towards a more restrictive approach on privacy protection in the UK? - Verfassungsblog
Can a Machine Learn Morality? - NY Times
Facebook fed posts with violence and nudity to people with low digital literacy skills - USA Today
Records show the Indiana BMV has been selling people's personal information - CBS 4 Indianapolis
FTC Privacy Probe of Amazon Ring Puts Khan's Agenda in the Spotlight - The Information
FB lobbied over poll rules: Papers - Hindustan Times
Social war: Will the Online Safety Bill keep us safe? - Politics Home
Want more recommendations on what to read? Follow us on Twitter