Trying to work out poll anonymizing

[conuly]

Can you just click a number in the poll? Just pick the number that indicates how many people have already voted. You don't have to tell me you voted or who you are, I just want to see if there's any rhyme or reason at all to how the numbers sort after the fact.

Poll #21816

This poll is anonymous.
Open to: Registered Users, detailed results viewable to: All, participants: 108

How many participants did this poll have BEFORE you voted? Only answer this one if the number is under 20

View Answers
Mean: 9.67 Median: 10 Std. Dev 5.75

0		1 (4.8%)
1		1 (4.8%)
2		1 (4.8%)
3		1 (4.8%)
4		1 (4.8%)
5		1 (4.8%)
6		1 (4.8%)
7		1 (4.8%)
8		1 (4.8%)
9		1 (4.8%)
10		2 (9.5%)
11		1 (4.8%)
12		0 (0.0%)
13		1 (4.8%)
14		2 (9.5%)
15		1 (4.8%)
16		1 (4.8%)
17		0 (0.0%)
18		2 (9.5%)
19		1 (4.8%)
20		0 (0.0%)

Okay, same deal, but this time it is between 21 and 41....

View Answers
Mean: 31.30 Median: 31 Std. Dev 6.14

21		1 (5.0%)
22		1 (5.0%)
23		1 (5.0%)
24		1 (5.0%)
25		0 (0.0%)
26		1 (5.0%)
27		1 (5.0%)
28		2 (10.0%)
29		0 (0.0%)
30		1 (5.0%)
31		2 (10.0%)
32		0 (0.0%)
33		1 (5.0%)
34		0 (0.0%)
35		2 (10.0%)
36		0 (0.0%)
37		2 (10.0%)
38		1 (5.0%)
39		1 (5.0%)
40		1 (5.0%)
41		1 (5.0%)

For everybody else, there is a ticky box

View Answers

TICKY
77 (100.0%)

Comments

[cosmolinguist]

So I was definitely User #1 when I answered the poll with "0," but now 0 is associated with User #2.

[conuly]

That's the sort of thing I was expecting, okay. So they're not just anonymizing, they're randomizing with each vote.

[jack]

And now it says:

User #4 -- 0
User #3 -- 1
User #2 -- 2
User #1 -- 3

I hadn't expected that. Would "reverse order" be consistent with the other poll you saw?

[cosmolinguist]

Even that's broken now.

User #5 -- 0
User #4 -- 1
User #3 -- 2
User #1 -- 3
User #2 -- 4

Maybe mixing up the numbers is part of the anonymization process? So you can't tell someone's answer even if you know when they answered?

[antisoppist]

Well I was the 8th and answered 7 but it makes me user 2.

User #8 -- 0
User #7 -- 1
User #6 -- 2
User #1 -- 3
User #4 -- 4
User #5 -- 5
User #3 -- 6
User #2 -- 7

It's got to be rearranging deliberately after each answer, hasn't it?

[jack]

Huh. Yeah, it must be anonymization, although I wonder if something's wrong with it.

[moxie_man]

Supposedly, #20 here. Q1 was if it was under 20. Q2 was if it was 21 or more. So, that left me with "Ticky". However now I see the results, I see you listed "20" as "under 20." I interpreted "Under 20" as less than 20.

[fred_mouse]

Ah! You are the reason the numbers didn't add up :)

But that implies a lot of people are missing the point of the ticky, because I saw 31, and there are 11 'ticky'!

[hudebnik]

After I answered 22, there were already 8 "ticky"s, several values in the 1-20 range that had no votes, and I think one value with two votes. Which I think says more about the people taking the poll, and/or the wording of the question, than about the anonymization algorithm :-)

[fred_mouse]

Interesting. There were exactly 21 responses to question 1, which is as I thought it should be. There were three zeroes, and three twos, which I figured probably had to do with people loading the post close enough together that one hadn't yet replied when the next one loaded it. However, the lack of the 20 being a ticky means that I was making unfounded assumptions!

And the pattern for question 2 is already that uneven (although at least there the 2s are following immediately by 0s, which fits my previously hypothesis better than the question 1 data does!)

wot I see

[fred_mouse]

question 1:
User #28 -- 0
User #27 -- 1
User #14 -- 2
User #3 -- 3
User #9 -- 4
User #12 -- 5
User #7 -- 6
User #6 -- 7
User #10 -- 8
User #25 -- 9
User #24 -- 10
User #4 -- 10
User #26 -- 11
User #22 -- 13
User #5 -- 14
User #8 -- 15
User #20 -- 16
User #17 -- 14
User #30 -- 18
User #13 -- 19
User #19 -- 18

question 2
User #21 -- 21
User #31 -- 22
User #15 -- 23
User #1 -- 24
User #32 -- 26
User #11 -- 27
User #2 -- 28
User #29 -- 28
User #18 -- 31

tricky
User #12 -- TICKY
User #25 -- TICKY
User #4 -- TICKY
User #5 -- TICKY
User #20 -- TICKY
User #17 -- TICKY
User #23 -- TICKY
User #21 -- TICKY
User #1 -- TICKY
User #19 -- TICKY
User #16 -- TICKY

Re: wot I see

[moxie_man]

Interesting. I only see totals for each choice. I don't get to see what each individual chose.

Re: wot I see

[fred_mouse]

Do you see links that say 'view answers' before the stripy red bars? Because that is what I clicked on to get the individual answers.

Re: wot I see

[moxie_man]

Until you pointed it out, no. They blend in on my screen as a header title in bold. I'm so use to links being underlined that if I see something in bold only, my brain interprets it as just bold and it never occurs to me to wave my mouse over it.

Re: wot I see

[fred_mouse]

Ah! I wasn't sure if they might be a paid account type thing, and thus literally not there, as opposed to secretly there but not coded as links.

[marahmarie]

Pollution: some of the "for everyone else" User #s are showing up under previous questions (unless the poll is literally inventing new numbers for them, at least for that question).

As far as inventing new numbers for the previous poll questions (the ones before TICKY) the first two comments seemed to show a shift-by-one trend, then it just aaaalll went off the rails, to my eye. Torn if that's an algorithm or an error, as someone suggested.

[conuly]

I think it IS inventing new numbers.

[wpadmirer]

It said 63 before I hit ticky, and 64 afterwards.

[marahmarie]

See, I was ready to conclude shift-by-one for the whole thing until I watched it go off the rails as it did a few times above. Either way I think it needs to be fixed, because to my eye it's not random enough (my best guess, though I could be wrong as DW's back-end tends to trip me up from time to time, is that whomever coded it intended shift-by-one and something-something-something code spaghetti happened unknowingly, so you get what we've got here).

The shift-by-one trend's the problem, though, as for it to be truly anonymized you don't want that happening at all (in its current state, which is an "at least pseudo-randomized mess", it doesn't help readers of the poll figure out much unless they really follow along obsessively, but it might give poll authors a few more clues than maybe they should have).

[flamingsword]

Unrelated to this post, but do you remember sharing a link about how to give Princess Buttercup agency in the plot of The Princess Bride by making her a poet? I am starting to think I hallucinated that blog post, because I have been looking for it all evening. Nada.

Please at least tell me if I am crazy.

[conuly]

Certainly I remember that link. If you give me a day, I may even be able to find it.

[flamingsword]

Oh thank you! I thought I had shared it with a friend whose special interest it is, but apparently I never did. My brain some days, I swear.

[elenbarathi.livejournal.com]

I was Participant 80.

[conuly]

Thanks. Looking at this, I think they're not just anonymizing, they're randomizing and switching up the numbers every time a new person votes.