Dataset availability

[gus-morales]

It would be nice to have a handle on the dataset, and see how each of the results compare along where you live, study, education level, etc. A friend and I would gladly submit a code with statistical analysis to work on something along these lines. So, is it feasible to add an option to download the latest dataset?

[ommirandap]

I strongly agree that there could be an interesting opportunity to collect important data here.
And it shouldn't be so difficult, after the poll, you can ask politely to the user if want to share their:
Age, Country, Gender (maybe), educational status, etc...
(actually, maybe you can find something interesting with google analytics data)

[TristanBomb]

While I agree this would be a great thing to have, especially to help balance the test, I'm not currently collecting data.

[Patrick330]

One solution would be to pull the server logs to see what final results people end up with, since they're included in the URL. Should be possible to post a text file of all the result URLs called and let people extract individual parameters for analysis.

Of course you'd have to accept some level of error from people like me who adjusted results until I got Nazism to scare my friends...

[mitar]

Ehm, what about those who do not want to have their data stored? Maybe a better way would be to ask at the end if they want to contribute their data to a common public dataset and in that case data gets submitted to the server. Opt-in.

[gus-morales]

In that case I vote for an opt-out approach.

[mitar]

I really do think it should be opt-in. This is to empower people to make a decision. We have already enough tracing on the web and guessing of my preferences.

[gus-morales]

I agree with your attitude towards tracing, but I assume the dev is not looking to sell that data, or to participate in marketing. These would be used for simple statistics, which in turn would go back to the general public.

[mitar]

Still. Even if you do an academic research on human subjects, that has to be opt-in with full consent f a person. Otherwise you cannot publish a paper. If you really want to make this dataset full useful, we should follow similar guidelines.

[Patrick330]

I think you misunderstand. When you visit websites your URL requests and IP address are logged by the webserver. You can use VPNs to anonymize your IP, but you cannot avoid having your requests logged because the server must know what pages to send you.

[gus-morales]

Also, there are tons of useful datasets that do not follow similar guidelines. Similarly, the opposite is also true. Tracing has more to do with ethics, not with how useful the dataset is. And website cookies do this anyway already.

[mitar]

Not necessary, if it is single page app where client-side changes the URL. This seems to be the case of this site at the moment. Furthermore, it is hosted on GitHub, so any URL requests logs are on GitHub.

[Patrick330]

Yes, that's true.

[mitar]

So we would on purpose want to change this so that users would be more traceable? Why would that be a good thing. On the other hand, we could make it opt-in, when JavaScript submits the results to a server for those who do want to participate in the research. I have done similar things in the past and many users at the end do like to contribute and help, especially if you tell them that they can also get the data. I think this is much better proposition.

[gus-morales]

Because to me (the guy who originally proposed to work on this dataset), this is not about tracing people. My code would not trace anyone nor I would sell the data to third parties.

Now, your second argument seems more appropriate to the discussion. I was suggesting an opt-out approach which I assume would yield more data (e.g. Dan Ariely's TED talk on decision making), but of course I am open to personal experiences on that matter.