Samaritans Radar

» see samaritans.org as it lived, on the wayback machine

~ the obituary ~

Samaritans Radar was a british product built around a british institution. The Samaritans have run a phone line for people in crisis since 1953 and the name carries real weight there. None of that context traveled, and in turkey the whole episode is a story you read rather than one you lived through.

It launched on 29 october 2014, built for the charity by a digital agency called Jam. It was free. You authorized it against your twitter account and it began reading the public posts of every account you followed. When a post matched its list of phrases it emailed you to say a friend might be struggling.

The intent was not in question then and is not now. Young people in difficulty do post about it, and often the post is the only signal anybody gets. The Samaritans built the most direct thing that could be built from that observation, and the direct thing turned out to be the wrong shape.

The gap is consent. The person doing the monitoring opted in. The person being monitored was neither asked nor notified, and the only way out was to send the charity a direct message asking to be added to a list.

~ the case file ~

~ what it actually did ~

The matching was keyword work, not judgment. Phrases like help me and hate myself sat on the trigger list. So did mentions of depression and any request for someone to talk to. Any post that hit one went out as an email alert to whoever had subscribed to that account, with no context and no way to tell a bad night from a song lyric.

By 4 november about three thousand people had turned it on. Because each of those subscriptions scanned an entire following list, the number of accounts under watch was very much larger. Nobody in that larger group had agreed to anything. Twitter's api made all of this trivially possible in 2014, which is a fact about the api rather than about the charity.

~ the nine days ~

The objection came from the people the app existed for. Mental health service users pointed out what the tool looked like from the other end: a stalker or a persistent troll could subscribe to their target and receive a notification at the exact moment that person was least able to handle it. Others said they would simply stop posting honestly, which removes the signal the whole thing depended on.

Lawyers raised the Data Protection Act 1998 and complaints went to the Information Commissioner's Office. On 7 november 2014 the Samaritans suspended it. The statement said they had listened. It also said the criticism had come as a surprise, which was the more revealing half.

~ what carried over ~

It was closed permanently on 10 march 2015 and the collected data was deleted. Nine days of running, then four months of quietly deciding not to switch it back on.

The residue is a design lesson that gets relearned every few years. A monitoring feature has two users, the watcher and the watched, and only one of them was ever shown a consent screen. Everything built since on top of a public social feed for someone's own good runs into the same wall, usually with less honorable intentions than this one had.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »