Thread · 9 tweets · 16 Nov 2020

When looking at a product the question to ask is: “Could I become *good* at this?” If the answer is “yes” then you’re looking at a tool. If not, you’re facing the modern equivalent of a TV channel: your brain is supposed to be off.
↺ 3
Companies don’t do this from malice, they do it because it’s easy. Run some A/B tests, tweak some hyperparameters: watch engagement go up and down! Tools on the other hand require thinking and (*shudder*) listening to actual people rather than just measuring them. Ewww!
↺ 1
To build tools you also need to hire designers (actual designers, not just to make things pretty). That’s hard too — much harder to know how to spot a good one than it is to throw engineers or data scientists at a project.
The typical example of this logic is Netflix. You can’t be good at Netflix. You can only vaguely hope that the random list of categories it throws at you tonight is less baffling than last night’s. It’s the quintessential “computer said no” bureaucracy: meaningless, all-powerful.
I say Netflix, but the other video services are just as alienating. Same in music: Spotify has a million ways to suggest music, but things you control, like playlist management, is entirely bad. Google News is in that same vein, and that’s also where Google Search is nowadays.
I say this is done without malice, but that only holds so long. The oppressive impact on individuals and the societal harms at scale are starting to be really well known: ignoring them is negligent.
↺ 1
I often speak with engineers who are baffled that their latest project is bad, and who don’t see how to change tack. The way I see it: build tools for people to do something *with*; not services to do things *for* them. Put differently: build for hackers, not for users.
↺ 6
If your reaction to that was “but they’ll never know how to use it!” then I would kindly, but firmly, encourage you to spend some time thinking about whether that arrogance is justified.