Oh, that's a really good point. For the product analyst job, we're hoping to hire someone who's already done research on online harassment, and can help us to learn from other people's approaches.
Your idea for using aggression/harassment scores in admin applications is really interesting; I hadn't thought of that before. Nothing's actually planned right now, just research and conversations, but it's neat to see people already coming up with interesting suggestions. :)
On Fri, Jan 27, 2017 at 9:17 AM, Lodewijk lodewijk@effeietsanders.org wrote:
Thanks Danny for the elaboration.
I don't want to contest the value of this work at all - sorry if that seemed implied. I think it's an effort that may be quite necessary - especially in some communities.
The set of tools you're describing to be developed, seem all to be related to a process that eventually leads to blocking people off our sites. That is what triggered my response. This process may be necessary in a number of cases (unfortunately), and helpful for the community health. But it is all 'after the fact' - once harassment has taken place.
What I am curious about, is whether there are also efforts ongoing that are focused on influencing community behavior in a more preventive manner. I'm not sure how that would work out in practice, I don't have the solution (although some ideas have been bouncing around). This work seems related to bullying in general - which happens unfortunately in schools and communities around the world - and research on this topic may help identify methods that could have a preventive effect. I have yet to see a 100% effective program, but it may strengthen the efforts for a healthier community.
I can see that where these approaches are still investigated, or non-technical, the community tech team may be less suitable for implementing them. But I do want to express my hope that somewhere in the Foundation (and affiliates), work is being done to also look at preventing bullying and harassment - besides handling it effectively. And that you maybe keep that work in mind, when developing these tools. Some overlap may exist - for example, I could imagine that if the harassment-identificationtool is reliable enough, it could trigger warnings to users before they save their edit, or the scores could be used in admin applications (and for others with example-functions). A more social approach that is unrelated, would be to train community members on how to respond to poisonous behavior. I'm just thinking out loud here, and others may have much better approaches in mind (or actually work on them).
Hope that clarifies a bit,
Best, Lodewijk
2017-01-27 17:24 GMT+01:00 Danny Horn dhorn@wikimedia.org:
The project has four focus areas, and blocking is just one of them.
Here's
the whole picture:
- Detection and prevention: Using machine learning to help flag
situations
for admin review -- both text that looks like it's harassing and aggressive, as well as modeling patterns of user interaction, like
stalking
and hounding, before the situation gets out of control.
- Reporting: Building a new system to encourage editors to reach out for
help, in a way that's less chaotic and stressful than the current system.
- Evaluation: Giving admins and others tools that help them evaluate
harassment cases, and make good decisions.
- Blocking: Making it more difficult for banned users to come back.
We'll be actively working on all four areas. There aren't a ton of
details
right now about exactly what we'll build, for a couple reasons. The
product
manager and the analyst haven't started yet, and the research that they
do
will generate a lot of new ideas and insights. Also, we're going to work closely with the community -- talking to people with different roles and perspectives, and making plans in collaboration with contributors who are interested in these issues. So there's lots of work and thinking and consulting to do.
But here's one idea that I'm personally excited about, which I think
helps
to explain why we're focusing on tools:
Right now, when two people end up at AN/I, the only way to figure out
whose
version of the story to believe is by looking at individual, cherrypicked diffs. You can also look through the two editors' contributions, but if they're both active editors and the problem has been going on for a
while,
then it's very difficult to get a sense of what's going on. Sometimes it really matters who did what first, and you have to correlate the two contributions logs, and pay attention to timestamps.
The idea is: build a tool that helps admins (and others) follow the
"story"
of this conflict. Look for the pages where the two editors have
interacted,
and show a timeline that helps you see what happened first, how they responded, and how the drama unfolded. That could reduce the time cost of investigating and evaluating considerably, making it much easier for an admin or mediator to get involved.
There are lots of UI questions about how that would work and what it
would
look like, but I don't think it would be too difficult on the tech side. The information is already there in the contributions; it's just
difficult
to correlate by hand.
Assuming it works, that tool could have a lot of good outcomes. Admins would be more likely to take on harassment cases, because there'd be greater return for the time investment. It would take some of the burden off the target, so they don't have to figure out which individual diffs they should provide in order to make their case. Also, it would be harder for harassers to get away with mistreating people, because they wouldn't
be
able to hide behind a smokescreen of random diffs.
As folks on this thread have said, there are lots of other components to tackling the harassment problems. There will probably be groups of admins and others who are especially interested in helping with the reporting
and
evaluation, and the Foundation could provide trainings and resources for those groups. Making changes to the reporting system will involve a lot
of
community discussions about policies and competing values. Some of those conversations and plans will probably be led by the Foundation, and some
of
them will arise naturally within the community.
For this specific team -- the Community Tech product team, working with
the
community advocate -- our focus is on doing research and building tools that will support those conversations and plans. We're not going to take over the community's proper role in setting policy, or making decisions about how to handle cases.
To Fæ's point, the community will determine the social and cultural decisions about how to treat harassment cases, and our team's job is to build software that will help to put those decisions into practice.
On Fri, Jan 27, 2017 at 3:06 AM, Fæ faewik@gmail.com wrote:
On 27 January 2017 at 09:21, Lodewijk lodewijk@effeietsanders.org
wrote:
...
Do I understand correctly that this particular initiative will focus
on
fighting harassment, and not necessarily on preventing it? Basically
in a
similar pattern that vandalism is fought on most wikipedia projects?
I really hope that prevention, education and (social) training will
become
a major point in the overall agenda, but I can imagine that we can't
pay
all that from the single grant :) So I just would like to place it in
the
proper context.
Best, Lodewijk
+1 Spot on.
The plan appears to hinge on blocks as the outcome. Based on cases of long term harassment targeted at individuals which invariably involved off-wiki doxxing or contacting friends and family members of their target, blocking Wikimedia accounts is an approach that may remove Wikimedia projects as a platform but does little to help reform the person causing harassment. I would rather see systems that include reaching out to the apparent harasser to help them recognize and deal with their anger or obsessive issues. Treating badly behaved individuals as the "other", without aiming for a lasting resolution, means we are back to the old days of telling the unfortunate target/victim to change their identity or grow a thicker skin as the on-line harassment may never stop.
Fae
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/ wiki/Mailing_lists/Guidelines New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/ wiki/Mailing_lists/Guidelines New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/ wiki/Mailing_lists/Guidelines New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe