When you don’t have enough information

11-Aug-2026

One of the things I like about data science is that there is usually a point where you can say: I don’t know yet.

You can collect more data. Run another experiment. Check the assumptions. Look for confounders. Try to reproduce the result.

There is something oddly comforting about having a process for not knowing.

Job applications don’t really work like that.

I found this out again today.

I was rejected for a PhD position I really wanted. It was in a research project where I had previously applied for a PhD position. That time I made it through the interview and mini-project to the final three candidates. I wasn’t selected; they chose an internal candidate.

It was disappointing, but I understood it. I had made it a long way through the process, and someone else was chosen. That’s how recruitment works.

The professor then encouraged me to apply for another PhD in the same project when her co-professor advertised one later in the year.

So I did.

This position was actually a closer match for my background. It was more machine-learning focused and much closer to the work I’ve been doing. I already knew something about the project, had previously demonstrated that I could get through the selection process, and had been explicitly encouraged to apply.

This time I didn’t make it through initial screening.

The rejection email was a standard template default: polite, brief, and containing precisely enough information to tell me that I had been rejected and absolutely nothing to tell me why.

And I find that surprisingly difficult.

I’ve been rejected for positions before. Many times. I know, intellectually at least, that not getting a job doesn’t mean I am bad at this, and that there are usually far more excellent candidates than positions.

But this one hurts.

I think part of it is that this wasn’t an entirely hypothetical possibility. I’d already been through the process once. I’d reached the final three. The professor had asked me to apply again. I thought I had a reasonable basis for believing that this might work.

And now I have to reconcile that with apparently not being strong enough to make it past the first screening this time.

I don’t know what changed.

Maybe the applicant pool was much stronger. Maybe they were looking for something specific that wasn’t obvious from the advert. Maybe my application did a terrible job of explaining why I was a good fit. Maybe there was a qualification or experience that screened me out. Maybe other people’s backgrounds were simply a much better match.

All of those are plausible.

I have no way of knowing which one is true.

Normally, this is exactly the sort of problem I find interesting. Something doesn’t add up, so I want to find the missing variable. What changed? What am I not seeing? Where is the discrepancy?

Unfortunately, there isn’t a dataset here. There is just a generic rejection email. And like most humans, I am very good at filling in the gaps, with something because a void doesn’t sit well.

I could decide that they didn’t really look at my application. I could decide that I wasn’t good enough. I could decide that the previous application was a fluke. I could decide that I should change everything about how I apply for jobs.

Any of those stories might be true.

None of them is supported particularly well by the evidence I have.

That distinction is much easier to maintain when I’m looking at somebody else’s data. It is considerably harder when the data point is me.

So perhaps there are two things to learn from this.

The first is practical. I’ll look at the application again when I’m less emotionally invested in it. I’ll try to work out whether there are things I could have communicated better. If there is an opportunity to get actual feedback, I’ll take it.

But I don’t want to turn that into retroactive certainty. Improving the next application is useful. Deciding that I now know why this one failed is not.

The second is less comfortable.

I need to get better at sitting with uncertainty when there isn’t a way to resolve it.

I don’t have to decide that the rejection means nothing. It clearly means that, for whatever reason, I wasn’t selected.

But I also don’t have to turn that single outcome into a story about my ability, my suitability for research, or whether I belong in this field.

I can be disappointed without knowing exactly why it happened.

I can learn what I can from it without pretending I have learned everything.

And I can admit that, for now, I still don’t know.

There is something rather annoying about discovering that one of the skills I thought I was reasonably good at — being comfortable saying I don’t know — is considerably harder to practise when the thing I don’t know is about me.