zlacker

[parent] [thread] 9 comments
1. ineeda+(OP)[view] [source] 2023-07-02 08:48:38
<*> was my own ad hoc convention in the moment. My “.” style is the default for the mobile phone I’m on, which is the US, and I don’t mind sharing that since I’ve mentioned it in plenty of prior posts. But I do honesty appreciate the analysis, those aren’t things I’d specifically considered when posting now or in the past, and it’s always useful to know what subtle signals can be picked up in that sort of detail. Heck someone observant could probably infer broad geo region (time zone) just from the fact that I am making these comments at this time.
replies(2): >>fho+ic >>correl+Ud
2. fho+ic[view] [source] 2023-07-02 11:17:26
>>ineeda+(OP)
Writing style is also a lot more individual than people recognize.

Iirc word histograms almost uniquely identify authors. Of course this is on larger amounts of text, but I guess you could identify users over seperate platforms this way.

E.g. Intend to use ellipsis (...) to separate thoughts in online conversation a lot. But I try to not do that in reddit, where I try to stay somewhat anonymous.

Still, I assume that it would be possible to correlate my reddit and HN account just by comparing the word histograms (ie which words I use and how often).

replies(1): >>ineeda+u02
3. correl+Ud[view] [source] 2023-07-02 11:37:13
>>ineeda+(OP)
Then you'll probably appreciate the following info. You have commented excessively on HN over the years. If you have done any publicly accessible writing with your real name attached, then it is quite easy to find out who you are by correlating writing styles. (If you don't have done so publicly then at least your employer will be able to do that using all internal writing on one hand and all your HN contributions on the other.)

Sounds far fetched, but it's really not that hard. Quite recently somebody hacked this for correlating HN accounts with each other and found alt accounts of people with high accuracy. Which people confirmed. And that wasn't even a serious attempt, just a little hack on a sunday night.

In a sense, it's all too late now since all your writing is already out there. But could be good to know for the future.

replies(2): >>sveron+bj >>ineeda+Xt
◧◩
4. sveron+bj[view] [source] [discussion] 2023-07-02 12:21:48
>>correl+Ud
It's unclear to me what you're meant to do about this besides never post anything, which is depressing. Hence I don't even try.
replies(2): >>fho+Pk >>correl+gL2
◧◩◪
5. fho+Pk[view] [source] [discussion] 2023-07-02 12:37:04
>>sveron+bj
I guess you can try to develop two very distinct writing styles.

Or, as I do, consider everything posted to HN to be linked to me. My handle is actually an abbreviation of my full name. I consider HN to be "professional" correspondence.

replies(1): >>balder+FH
◧◩
6. ineeda+Xt[view] [source] [discussion] 2023-07-02 13:48:25
>>correl+Ud
All true, but that’s a level personal detail I’m prepared to live with having it out in the world. There’s a limit to how much mental energy I’m prepared to put into worrying about doxing, and it doesn’t go this far. But it is a remarkable aspect of just how much we can be fingerprinted by so many different things we do in life.
◧◩◪◨
7. balder+FH[view] [source] [discussion] 2023-07-02 15:26:18
>>fho+Pk
You can have LLMs rewrite your texts.
replies(1): >>tjoff+VP
◧◩◪◨⬒
8. tjoff+VP[view] [source] [discussion] 2023-07-02 16:10:58
>>balder+FH
And we are back to the depressing part where you can't even bother.
◧◩
9. ineeda+u02[view] [source] [discussion] 2023-07-03 00:31:22
>>fho+ic
Yes, some of my academic work in comp-ling (massively outdated by today’s advances) explored things like the perplexity scores of different Shakespeare plays to explore the controversial claim that some the work attributed to him was actually done by Marlowe.

(As a complete aside, that program of study also included Forensic Linguistics which truly fascinating. And of course the work of Claude Shannon and information theory, though not in any great depth)

◧◩◪
10. correl+gL2[view] [source] [discussion] 2023-07-03 08:18:20
>>sveron+bj
One option is to change HN accounts often. Every few posts you need to make a new one. Successfully correlating writing styles needs some data and if there are only a few sentences to go off then that's not enough.

I know that the site guidelines discourage that, but what can you do.

[go to top]