Writing action items a scholar can actually act on

Look at the actions from your last ten supervision meetings. If most of them start with a vague verb — explore, look into, think about, continue — you have found the reason your scholars keep arriving with less progress than either of you expected.

This is not about micromanagement. A doctorate is training in independence, and a supervisor who specifies every step is failing at the actual job. But there is a difference between leaving the method open and leaving the task undefined, and supervisors conflate the two constantly. "Work out how to handle the missing data" leaves the method open. "Look into the missing data" leaves everything open, including whether anything needs to happen at all.

The test: could someone else tell whether it was done?

The single most useful check on an action item is whether a third party, with no context, could look at it in a month and say yes or no.

• Look into alternative clustering approaches — no. There is no state of the world that clearly counts as having done this.
• Run k-means and DBSCAN on the pilot set and bring the silhouette scores to the next meeting — yes. Either the numbers exist or they do not.

Notice the second version does not tell the scholar how to do anything. It does not say which parameters, how to handle outliers, how to interpret the result. All of the research judgement is still theirs. What it fixes is the boundary: they know what finished looks like.

This matters most for the scholars who are struggling, which is exactly backwards from how most supervisors apply it. A confident scholar will convert a vague instruction into a concrete plan themselves. A scholar who is lost cannot, and a vague instruction leaves them staring at it for a fortnight, increasingly unable to admit they do not know where to start.

Four things every action needs

A verb that names a deliverable. Run, write, draft, send, read and summarise, tabulate. Not consider, explore, think about, look into. If the verb does not produce an artefact, there is nothing to check.

A scope boundary. Read three papers, not read the literature. Draft the methods section, not start writing up. Unbounded scope is the most common reason a task stalls: the scholar cannot start because starting means committing to an amount of work they cannot estimate.

A date. Not because the date is sacred, but because a task without one is not scheduled against anything. A scholar with six open actions and no dates will work on whichever feels most urgent, which is usually the one they already know how to do.

A named owner, when there is any ambiguity. Some actions are yours. "I will email the ethics committee" belongs in the record exactly as much as the scholar's tasks, and writing it down is how you find out that three of the last five blockers were waiting on you.

That last one is uncomfortable and worth sitting with. When supervisors start recording their own actions in the same list, a common discovery is that the bottleneck was never the scholar.

Three or fewer

A meeting that generates seven actions has generated zero. The scholar leaves, looks at the list, cannot tell which matters, and picks by comfort rather than importance.

Three is a good ceiling for a fortnight. If a meeting produced more, the honest thing is to choose: which of these actually has to happen before we next speak, and which am I listing because it occurred to me? Move the rest to a backlog you both look at occasionally. They are not lost, and they stop competing with the work that matters now.

This is also a useful signal about the meeting itself. Seven actions usually means you covered four topics lightly instead of one properly.

Say the deadline is soft when it is soft

Academic deadlines come in two kinds and supervisors routinely fail to distinguish them. The ethics submission date is real. "Have a draft by the end of the month" is usually a pacing device.

A scholar who cannot tell the difference will either treat every date as immovable — and burn themselves out, or quietly hide a slip — or treat every date as advisory, including the ones that are not. Both are worse than being told.

It costs one clause: the conference abstract is a hard deadline, the draft methods section is so we have something to talk about, and if it slips a week that is fine. Scholars remember this kind of thing, and it buys you a great deal of honesty about the deadlines that actually matter.

Review the list before you write a new one

At the end of the meeting, before you write the new actions, read the old ones out. This takes ninety seconds and does two things.

It catches the action that has now been overtaken by events — you decided to drop the survey, so the survey protocol task should die rather than sit on the list making the scholar feel behind. Stale tasks are corrosive out of all proportion to their number.

And it surfaces the pattern. An action appearing for the third time is information, and the right response is almost never to write it a fourth time. It is to ask what is actually in the way — and then to rewrite the task smaller, or take it off the scholar entirely.