Hacker News Reader

Comment by Const-me | Hacker News Reader

Very likely you’re right about the current state of alt right movement (I live in Europe and haven’t been monitoring the subject).

But GP only mentioned alt-rights as an example of where might the voters turn their preferences.

Even at Google where James worked, many employees agree with his memo (36%), and the majority, 54% of them think the memo is harmless and was OK to share: https://basicgestalt.wordpress.com/2017/08/06/press-f-for-ja...

After Google fired him, these 54% will inevitably think about what other ideas they think are harmless might cause them to be fired. Not exactly comforting thoughts.

View on HN

For the life of me, I cannot find a legitimate source for that poll. I see the image popping up on a lot of less-than-reputable websites (wordpress blogs, tweets, businessinsider, etc) but no actual source for it, let alone any idea of its methodology.

Since it's allegedly Google internal poll, you probably won't find more information on it beyond screenshot. After what happened to Damore and calls for his supporters to be fired as well, most people are probably clamming up now.

I can't say that inspires confidence that the poll is even legitimate. Do we at least have proof that it actually was an internal Google poll and not someone's 15 minutes in photoshop?

No, we don’t have that proof.

However, the design matches the defaults of google forms, exact colors and their sequence, too: http://imgur.com/WPC0ogZ

Also the header says "gd-jpeg v1.0", it’s a cross platform open source C library with bindings for all modern programming languages including Go and Python: https://en.wikipedia.org/wiki/GD_Graphics_Library

Even if that’s a fake, it’s a good one.

I mean, GD has been around for over 20 years. It's not obscure. It's extremely popular. Most coders have used it at some point.

I could whip upa similar graph without breaking a sweat that says whatever I'd like, and it would have all the same hallmarks of this one. I could even do it in PHP using ImageMagick (if I felt masochistic) and fake the file header to say it was created using GD. Or build it in photoshop, and pull it into a binary editor.

So I think unless there's someone who can point to greater evidence that this is an internal graph, there's just no reason to believe it is.

> So I think unless there's someone who can point to greater evidence that this is an internal graph, there's just no reason to believe it is.

So I think unless there's someone who can point to the actual internal graph, there's just no reason to believe this is fake.

The distribution this poll shows is not so outlandish as to be unbelievable, and importantly it does not lend itself to overwhelmingly support one side over the other. Why would anyone spend time and effort to fake it?

> Why would anyone spend time and effort to fake it?

It's a heated argument on the internet. There'd be plenty of incentive to fake a graph in order to bolster one's opinion.

My point is that if you are going to fake a graph to bolster your opinion, you might as well fake one that actually bolsters your opinion, instead of just mildly suggesting that not everyone disagrees with you.

I'm really disappointed we're even having this conversation on Hacker News. A graph with no source, no methodology, no context is useless. We're computer programmers, we should know enough about data to recognize this and dismiss it until we have those things.

Assuming it was legitimate, 282 self-selected responses from a company with tens of thousands of employees wouldn't pass even the most basic "is this poll statistically valid?" test.

> wouldn't pass even the most basic "is this poll statistically valid?" test

It totally would. Statistically, a sample size of 282 out of a population of 60000 is pretty good.

If this test was authentic, we would be 95% sure that the true percentage of the population who think that the memo is harmless is between 47% and 60%.

Presumably that would require a population-representative sample of 282 though, not a self-selected (and therefore potentially biased) sample.

Otherwise you could run a poll "Who thinks [minority] should leave the USA?", get a high "yes" response, say "THE POPULACE AGREE!" but then discover only racists/whatever have replied because Breitbart/4chan dogpiled the poll - which would make it 100% invalid, no?

Reply on news.ycombinator.com