The running app on my phone holds every run I have ever logged, mapped to the metre, and has never once suggested what to run tomorrow. The contacts app holds years of messages with people I love and watches the oldest threads go quiet without comment. The sports app counts goals and wickets on the far side of the world and will happily spend a night of my sleep on any of them, because it prices its own alerts at nothing. And a government office somewhere holds the exact dates my rego, my warrant, and my rates fall due, and leaves the remembering to me. In any given week the four of them measure almost everything I do. Not one of them, ever, decides anything.
You could call this modesty, and mean it as praise. A tracker is not a coach. Deciding what to run tomorrow, who to ring, what is worth being woken for, is my call, and I would resent software that presumed to make it. The numbers are laid out cleanly and neutrally, and leaving the judgment to me is the respectful thing to do. Anything else would be a machine telling a grown adult how to live.
I believed that, and “the numbers are neutral” turns out to be the exact claim a long line of careful work says is false.
What a measurement leaves out
To measure a thing you first have to make it legible, and that is not a neutral act. James Scott spent Seeing Like a State, in 1998, tracing what happens when an institution standardises the world so it can be counted: it captures what fits on the form and discards what does not, and what does not fit is usually the local, tacit, adaptive know-how that was making the thing work. (James C. Scott 1998) Scott called that discarded knowledge mētis, and his point was that legibility does not observe from the side. It reshapes what it looks at, and it quietly throws the judgment away.
Jerry Muller gave the managerial version in The Tyranny of Metrics: metric fixation, the faith that measured performance can stand in for judgment based on experience, because judgment looks subjective and a number looks solid. (Jerry Z. Muller 2018) Muller was worried about measurement replacing judgment. The tools on my phone have found a stranger destination. They do not replace the judgment with a number. They gather the numbers obsessively and then decline to make the judgment at all, and hand it back as though that were a courtesy. The running app did not neutrally leave tomorrow’s session to me. It measured everything about today so well that the one act it could not reduce to a chart, deciding, is the act it drops in my lap, tired, at the end of the run.
The same refusal, four times
Once you can see the move, it is always the same move. The running app could take today’s effort and shape tomorrow’s session around it, which is the only thing that would count as coaching, and it does the archiving instead. The sports app could judge whether one game, tonight, clears the bar for a night of my sleep, and instead wakes me for all of them or none of them. The contacts app could notice that one friendship, specifically, has gone quiet for months and that tonight would be a good night to ring, and instead offers a searchable history of the drift, a friendship measured like a sales pipeline. The agencies could sequence what I owe and prompt me at the moment I could act, and instead each posts its own due date and waits. Measure, then refuse. The tool captures beautifully and decides nothing, and the deciding, the only part that was ever hard, is handed back to the exhausted person the tool was supposed to help.
The half of the job that can be blamed
A fair objection lands here, and it deserves a straight answer, because the opposite failure is real too. If measuring everything and deciding nothing is one error, measuring everything and deciding everything is the other, and it is arriving fast: the first use of each new generation of machines is to bolt a recommendation onto every judgment we used to make ourselves. Markus Schuller made the finance version of the warning in April 2026: as tools shift from recording what happened to recommending what to do, people start accepting the recommendation without doing the reasoning, and drift into dependence on systems they no longer understand. So the answer is plainly not make the machine decide everything.
But that is not what any of these tools is being asked for. A narrow, mechanical band of judgment is genuinely low-stakes to automate and genuinely useful: turning today’s run into tomorrow’s session, deciding whether one alert is worth a night’s sleep, noticing a friendship has gone quiet, prompting an obligation at the moment it can be met. None of that is taste. None of it is a life decision. Each is the small, dull call that the measurement makes obvious and then declines to make. And the reason it declines is not respect for your autonomy. It is that the two halves of the job are priced completely differently.
Measuring is cheap, legible, and safe. It fits on a dashboard, it demonstrates activity, and no measurement was ever blamed for being wrong, because a measurement does not commit to anything. Deciding is the opposite. It is a claim about the future that can turn out wrong, and a thing that can be wrong is a thing that can be held responsible. Rory Sutherland tells the story of a travel company whose website converted at around 0.5% and whose phone line converted at close to 30%, and which poured its effort into the website, because the website was the part it could measure. (Rory Sutherland 2019) The same instinct runs through every one of these tools. They keep the half of the job that is easy to count and impossible to blame, and they give you the half that is hard to count and easy to blame, and they call the handover freedom. What I am left holding, at the end of the run, at three in the morning, months after the last call, is not data. It is the decision the machine measured its way right up to the edge of, and then left for me.