Comment by iamnothere | Hacker News Reader

Comment by iamnothere | Hacker News Reader

Parent Full thread

The database issues are a problem, agreed, hopefully they get it fixed. You shouldn’t publish something claiming to be based on data without making sure access to the data is available.

Also a great question about Kissinger.

> Before social media, these sorts of things would be reported in the school and perhaps also local newspapers.

Actual disinvitations, yes, but I imagine that it would have been harder to find examples of unsuccessful pressure. Not everything makes into the paper, depending on who the editor is, and an unsuccessful pressure campaign may not be considered a real story. I don’t see how that past data could even be collected now in a convincing way.

> Are you sure it was Pew?

I think you’re right, I have also read the FIRE article and I may have confused it with an earlier poll from Pew about political violence, along with a more recent one about free speech and the media. https://www.pewresearch.org/short-reads/2025/10/23/americans... https://www.pewresearch.org/journalism/2026/09/15/americans-...

With regards to past violence, I would say that the early instances were indicative of the later trend. But obviously not every trend continues forever (most don’t). I don’t think it’s something to panic over, but it shouldn’t be ignored.

That's rather the question, isn't it? Why do you think the present data collected by FIRE was done so in a convincing way?

I read the methodology and wasn't convinced that the collection isn't biased by the marketing push for awareness of FIRE's mission, or by changing methods in data collection.

Their graphs, for example, don't even show error bars, which makes it impossible to determine if the trend they report is meaningful or random noise.

Regarding old campus newspapers, part of the "expensive and slow scholarly effort" means also reading the followup papers to see if there is information about a meeting being held or canceled, and if there isn't, include that in the statistics collection.

I don’t think the methodology is perfect, but they have been reasonable in the past. Not as good as the EFF, and with a past rightward slant (but it’s encouraging to see them pointing out the right here). They are a useful signpost on civil liberties from the right much as the ACLU can be a useful signpost from the left.

I suspect that the data collection will improve from this point forward if they commit to this series. As I said, I don’t really think that it’s feasible to get convincing data on this particular issue from the past, even if you mine the newspapers and whatever fossilized meeting schedules can be extracted from the archives. You could craft a narrative around whatever you find, but I don’t think I’d put much stock in it, whichever way it goes.

As I said earlier, I’m just happy someone is collecting this data now moving forward. I think it’s an important indicator, and definitely something to watch in the future.

How do you know they have been reasonable? Because they come to conclusions you agree with, or because you have investigated their methodology and found it sound?

"If they commit to this series"? You know their data starts in 1998. How long until you think they are committed?

If there are no error estimates, how can you even tell if it's an important indicator upward or downwards?

> "If they commit to this series"? You know their data starts in 1998. How long until you think they are committed?

Ok, so database search does seem to work for me at least. I found a 2020 Kissinger cancellation but not the 2018 one. I am curious about its absence. There was only one NYU issue recorded that year. Looks like you might have found a missing data point.

Anyway, if they stick with the current methodology until 2030, I’ll say that’s a solid start. That’s 15 years since 2015, which is a reasonable arbitrary year past which the effects of modern social media would be in play. Trends should be very clear at that point, and data collection methods shouldn’t be too different from the start of that period.

As for why I respect their reporting, I read some about their methodology in the past, when I first started following them, and I also have read criticism of them, which I did not find compelling. I don’t generally dig into every little detail of it because I have trusted sources who trust them, and that’s how you generally find your way in a world with too much data and too little time. Regardless, although they are imperfect (that missing data point!), they seem to be mostly thorough, and reading through the methodology PDF it looks like a good overall approach.

Given that the first example I came up with is not in the database, that strongly suggests the dataset is not all that thorough.

That, combined with the lack of error bars (with estimates for possible increased signal sensitivity in their collection method), and the fact they changed methodology while their reports imply an unchanged methodology, does not lead me to the same conclusion you've reached.

Reply on news.ycombinator.com