Competition To Identify Sexual Predators In Chat Logs 273
An anonymous reader writes "Researchers from the University of Lugano, Switzerland, and other universities from the U.S. and Europe organize a competition to automatically identify sexual predators in chat logs. The task is described as: 'The goal of this sub-task is to identify classes of authors, namely online predators. You will be given chat logs involving two (or more) people and have to determine who is the one trying to convince the other participants(s) to provide some sexual favor. You will also need to identify the particular conversation where the person exploits his bad behavior.' Their data set covers hundreds of chat logs with dozens of true positives (i.e., chats where one is trying to hit on another)."
Ummm (Score:5, Insightful)
automatically identify sexual predators in chat logs
OK sounds good hate those guys
trying to convince the other partecipants(s) to provide some sexual favor
Whoa whoa who here, thats me and my wife not a bunch of predators.
Whoever would have guessed that the govt trying to get into my bedroom would mean they're scanning my chat logs.
I suppose Target store wants to monetize this too...
False positives? (Score:5, Insightful)
Dateline NBC (Score:4, Insightful)
But... (Score:3, Insightful)
What if I am just trying to get laid? Seriously, how does one determine from chat text whether a person is a 'sexual predator' vs. someone who is just looking for a casual hookup? Wouldn't the approach be similar if not identical? I smell a FAIL.
Of course, the results of this 'competition' will likely get support from conservative, big brother regimes as a way to ring up innocent and horny people - particularly targeting young men - online.
Gender based stereotyping, convictions and punishment coming soon to the interwebs and country you live in. That's just wonderful!
Comment removed (Score:4, Insightful)
Re:Ummm (Score:4, Insightful)
Do try to differentiate between a computer science exercise and a government policy.
Competition to help sift all communications? (Score:5, Insightful)
While cloaked in "won't somebody please think of the children" language, it appears to me that this project is really all about developing technology to rapidly scan a mountain of text conversations to identify any instances of behavior for which you have a few documented prototypes.
Swap in political activist, opposition party, occupy movement, flash mob, or hackers, and the project doesn't seem so appealing. The goal sounds like they would like to find an engine to which you could feed in a few examples and have a few thousand computers watching all conversations on the net.
Why would Universities participate in that? Are these people that naive? Why not spend the money on education materials, or web sites explaining the sexual predators techniques so at risk populations can be smarter, rather than helping governments build Skynet?
Re:This is So Easy (Score:4, Insightful)
Facebook.
Sex is bad (Score:4, Insightful)
First, "being persistent" (common dating advice for men in earlier decades) became "stalking". Telling a woman she looked hot became "sexual harassment", even when the man had no power over the woman. Now, asking for sex makes you a "sexual predator". And if a woman agrees to sex, men have to worry that she may later claim she was raped.
The "sex is bad" side has won.
Re:Ummm (Score:5, Insightful)
Do try to differentiate between a computer science exercise and a government policy.
Do try to be a little less naive.
Just scroll down a few stories here on Slashdot and find a UK proposal to scan ALL internet communications in real time [slashdot.org].
Re:Obvious? (Score:5, Insightful)
It isn't always obvious because efforts to get into someone's head aren't always obvious.
Some people will attempt to groom a chat participant - they will ask more or less innocuous questions, but occasionally throw in one that is just a shade less innocuous than the others. Over time they will push the limits, until eventually some of the most outrageous stuff seems like it's just par for the course from this person. They'll couch all of this in the guise of being a mentor or friend, will back off if their target gets a little iffy, but will try to reconcile and take another tack. When this method is used against a target from a vulnerable population (kids, for example) it's scary how effective it can be and how easily even people who are not in vulnerable groups get taken in (see: people who fall for scams).
Eventually predators will shift to a more active part once their target has been willing to talk openly about previously forbidden subjects, and they'll attempt to get a cam session, phonecall, pics, meetings, whatever. The target might agree to go on cam or to pics, and at that point the predator has them - "Hey, if you don't keep doing this I'll post those pics/videos everywhere" etc.
It's easy to recognize the obvious and unsubtle ones, but it's a lot harder to recognize (from a relatively small sample) the more crafty ones unless you're an outside observer. For example, if I were someone being groomed by a predator as I described, I might not balk at a question as to whether I had pubic hair since I'd already been conditioned to trust that person. But if I were an outside observer it would be obvious that is untoward - unfortunately for many people targeted by predators, no outside observer is there to kind of make them realize what's going on.
As a researcher who often works with young people who have been exploited or put at risk, I've been given chatlogs from predators like the ones I've described above and was just astonished at how things progressed. However, quite a few of the skeevy questions that were asked by predators were ALSO asked in completely innocuous relationships and in that context were not nearly as skeevy. Just flagging based on questions or terms isn't enough - it's a context that needs to be understood.
The goal, I imagine, in the case of this contest, would be to help automate the process of that "outside observer" to have the software check for suspicious behavior/history and throw a flag once it passes a certain threshold but BEFORE the target gets exploited, and possibly to minimize the number of false positives so that extensive resources aren't wasted on non-predatory relationships.
Ethical considerations aside, it's an interesting problem and could be applied to a number of areas where you're attempting to detect non-obvious manipulative behavior in any kind of multi-party interaction.
Re:Ummm - NO! (Score:4, Insightful)
And of course, its always more palatable to couch surveillance projects in "Somebody Please think of the Children" language.
Re:Sex is bad (Score:3, Insightful)
There is a difference between being persistent and stalking a woman. If you don't know the difference you should probably not try to be persistent.
Choose who you flirt with wisely and you won't have most of these problems. In an office environment let her flirt first. Then flirt back, but don't escalate and always have an out. You should be doing that anyway since you don't want to be "too interested". When you compliment a woman remember she is more than just good looks. No one was ever sued for sexual harassment for complimenting a woman on a witty response or on a job well done. Save the sexual/physical compliments for your first date.
If you have consensual sex and she accuses you of rape, YOU picked a lousy woman. There were warning signs. You didn't pay attention to them. (She's a worthless slug of a human being, but how that information helps you after the fact is beyond me.) This happens because men are stupid and think "getting laid" should be a goal unto itself. Form a relationship with a woman first and this will never happen to you.
All these 'sex is bad' rules are really 'women feel uncomfortable' rules. The rules are stupid as hell. But as a man you shouldn't be doing things that creep women out anyway. If you do, you are doing something wrong - fix it!
Wait a second.... (Score:4, Insightful)
Since when is one person chatting with another and asking for "sexual favors" suddenly a sexual predator? If I'm hitting on a person and ask to see a naughty pic this is predatory behavior? If the person is of age and I'm of age WTF is the issue? If they asks me for a pic and I'm interested am I being preyed upon somehow? What if I welcome this, what's the issue? The assumptions here are tremendous IMO! Please tell me age plays SOME factor in all of this sheesh!
Re:what about (Score:5, Insightful)
Agreed, I'd feel a lot better if part of this competition was zero (not "acceptably low") false positives. Some backwards places in the world (yes, I am speaking specifically of America) being accused of sex crimes is a Bad Thing and will ruin your entire life, even if the accusation is baseless. It is not acceptable to create an algorithm that will ruin innocent people's lives with some probability, if used for its intended purpose.
Re:what about (Score:2, Insightful)
Thanks to the incomprehensible network of laws, chances are the victims of a false positive are already guilty of something else, so they deserve it.
And I guess America deserves it for continuing to vote Republocrat.
What's That Sound.... (Score:5, Insightful)
It's Alan Turing spinning in his grave.
Re:False positives? (Score:4, Insightful)
The cultural expectation in the West has been, for quite some time, that if you're a white male, you're a sexual predator. That's what Women's Studies has been teaching to Education and Business majors now for several generations, and is often considered a requirement (as I understand things). Many convictions today of sexual harassment, as well as rape trials, are based on that thin veil of sexism and racism.
Re:what about (Score:4, Insightful)
What, exactly, is wrong with trying to get sexual favours and how does that make one a "predator"?
Re:False positives? (Score:3, Insightful)
Many convictions of sexual harassment are based on actual sexual harassment.
Shouldn't it be "all convictions of sexual harassment are based on actual sexual harassment"? Or at least "most"?