Disruptive Proactivity

7th Nov Update: They’ve “made the decision to suspend the application at this time for further consideration“.

I’ve been asked many times what I think of #SamaritansRadar. There are many unanswered questions.

While my day job hasn’t take a position on this (I’m not sure the Samaritans have said enough publicly to substantiate an informed position), it does look at projects a set of criteria, which include: consensual, safe, and transparent. So some parts of that process can be applied here by me, without the rigour and diverse inputs that normally go into discussion documents that go through the process.

SamaritansRadar v1

Oh dear.

Oh dear, oh dear.

It’s sort of transparent, in that they’re being very clear how badly this is implemented, and continuing to dig. However, that’s not really what that means.

The spectacular failings of SamaritansRadar have been covered elsewhere. But in terms of consensual, safe, and transparent, it probably gets a “at least you tried” a bit for only the first.

It’ll need another iteration or three, but let’s consider what a version 2 might look like. They have to go somewhere from here, and their preferred direction currently seems to be down. There are other options.

SamaritansRadar v2: for tweeters, not to tweeters — a subset of considerations

Hopefully, in the next iteration, they’ll stop digging and at least acknowledge they’re hearing concerns and their effects.

They seem insistent on doing something, but hopefully it’ll stop getting worse. Whether that is a good idea is something for experts in the area to assess on a substantive proposal (which this is not). The current iteration has had used the public face of their head of policy, who is probably wishing he could hide behind a tree.

What do the counsellors and the core staff of the Samaritans think about how it will help their work? What are their user needs and stories? What will help people who need it, without affecting others?

But, for now, let’s limit this discussion to possible technical implementations of the current, publicly stated, seeming goal of the project: “allow a user to be told of keywords in the tweets of the people they care about.” That design needs iteration, but that can be considered a little later. Ending the current harm must happen before good can be considered.

If you’re going to mass surveil twitter, there are a few ways of doing it. The first, the current way, is an app that mass surveils all tweets of users who have given permission for the tweets they read to be scanned. This hides the tool behind the identity and permissions of those who use the app.

An alternate approach would be, have a @samaritansradar user and run a search. As a native of twitter, that user exists like all others, and can hence be blocked like all users. For those distributed blocklists of trolls used by some who have been victimised in the past, @samaritansradar can be added to that list. It will be immediately obvious to any user of twitter whether they’ve blocked the account, whenever they felt the need to look. Some control would be given back to those this app has caused distress, without any other changes. There need to be other changes, but the control is transparently within the control of the person the Samaritans are trying to help, and not a third party.

Further changes must take account of who uses the tool. At the moment, all the detail is hidden away on the servers of the media company that built it. Let’s just hope they don’t get bought out and that list monetised, but that’s again a different problem (needs solving, but is not a technical issue). Moving the blocking function to twitter would have removed all state from non-twitter servers. That seems important.

So, assuming that is the case, there becomes a question of what happens when something actionable has been picked up. How can the principles of safe, consensual and transparent be maintained there?

Firstly, has anyone else noticed the tweet? If you check in 15 minutes time, has anyone replied to the tweet, retweeted, favourited, etc. That likely suggests that the tweet may have been someone quoting a Judge, rather than a cry for help. If someone phones you up or sends you a direct communication, with a cry for help, that suggests it’s somewhat timely. As a cry into the ether, there can be a delay in response.

Secondly, the Samaritans are very fond of saying how they’re working with twitter, and have access to promoted tweets. Why don’t they use them? If you have someone who you feel may benefit from a reminder someone exists, twitter has the given the Samaritans the ability to inject that message into the relevant person’s twitter stream. Whether that’s consensual and transparent depends on the details. Also, the fact that it’s an advert, suggests that it may not be hugely successful. It needs to be done carefully, but it can be done.

Thirdly, if that hasn’t worked repeatedly, then, maybe, there may be a case for a clearly articulated workflow, which has adverse outcomes well considered, where the app (and, if the account is activated by following it, this is easy), sends users who care a DM pointing at the tweet. How “users who care” is defined, must include some analyses of recent past interactions, positivity, and potential for catastrophic fuckup. This is likely to involve a human intervention on the part of the international network of Samaritans organisations.

It would not be a net reduction in harm if someone were to say they were sad, only for the Samaritans to DM their recently-ex-partner to point that out, thinking they’re helping. A human intervention is necessary here.

This isn’t a proposal

The above isn’t a proposal for how the Samaritans should do something, but a recognition that the current implementation doesn’t necessarily live up to the standards they usually meet, and that alternates are possible. There are a couple of hard questions I entirely ignore above, but which are raised elsewhere. The options aren’t “this” or nothing, there are ways forwards which meet the needs of those whose mental state has been impaired by the current version, and those who people with good intentions were trying to help. That needs to be a conversation with multiple diverse views and iteration based on feedback, and not run by a poor imitation of an ostrich.

The Samaritans are an organisation whose mission is to listen. They are very similar to GCHQ in that way, although it does seem that the ad agency working pro bono might have got the two mixed up.

Whether the Samaritans Radar survives in operation any longer than the Samaritans Head of Policy is in post, remains to be seen. The Samaritans are currently recruiting a new “Head of Digital”. They really need it.