At 4:12 on Tuesday, a colleague asks you to join a meeting about a project that has gone wrong. You never see the invitation. Your calendar assistant rejects it because you told it to protect your afternoons for focused work.
On Wednesday, your colleague is distant. She thinks you chose not to help. Technically, the calendar did what you asked. But did you make the decision she received?
A preference is not every future answer
Blocking interruptions is sensible. Most meetings are not emergencies, and a tool that asks permission for every routine choice would barely save time. The whole appeal of automation is that yesterday's instruction can handle today's small decisions.
Yet a standing preference is a rough description of what matters to you. It cannot always tell the difference between a status update, a request for solidarity, and a conversation in which your absence will be understood as a statement.
The problem is not that the assistant made a mistake in reading your rule. It may have applied the rule perfectly.
Who should explain the refusal?
You could tell your colleague that software declined the meeting. That is true, but it risks sounding like an alibi. You chose the software, set the rule, and benefited from the quiet afternoon.
On the other hand, expecting you to inspect every automatic refusal defeats the point. A better system might distinguish routine invitations from unusual ones, send a digest of what it declined, and make exceptions easy to review. None of those safeguards can guarantee that it will recognize the human meaning of every invitation.
The small decision hiding inside the large one
Imagine the assistant instead replies: 'I am protecting this time. If the matter is urgent or needs this person's judgment, please say why.' That gives the sender a path back without pretending the calendar knows your priorities better than you do.
It also shifts work to the person inviting you. Someone with less authority may hesitate to insist, even when the meeting matters. Convenience is not distributed evenly.
Perhaps the useful question is not whether machines should make decisions for us. They already make countless minor ones when we set filters, rules, and defaults. The question is which decisions deserve to come back to a human before they become a message to someone else.
What is one choice you would gladly automate for yourself, but would hesitate to let a machine communicate on your behalf?
Part of Beyond the Known, an independent journal by Arin Vale.
Related reading: The Rule Still Running — when a useful setting outlives the choice behind it
Read our editorial approach

