Wellfare Check Algorithm

Community
1

Is there a feature, plugin, or system that can flag users for a welfare check? If a user’s posting, commenting, or even login and reading pattern changes, the system would flag them for a welfare check – just an outreach, by email or other means, to see if the user is okay.

This would be in the case of a community that has a well-established base of users, and a history user behavior. It could be an opt-in feature, where the user consents to be contacted if they disappear for awhile.

If someone usually logs in one average X times every Y days over a given period (months?) the system establishes what is the norm for that individual.

  • Aiko logs in, on average, once a day, and has done so for more than a year. So the welfare check would trigger if Aiko does not login for, say, 14 days.
  • Jasper logs in on average once per week, and has done so for more than a year. So the welfare check would trigger of Jasper does not login for, say 60 days.
  • Inés logs in on average twice per week, and has done so for 2 months. Inés does not meet the “community integration” threshold for a welfare check.
  • Zubair has bursts of activity, and then goes away for months at a time. There’s no established pattern, so Zubair does not meet the “consistency” requirement for a welfare check.

Here’s why I’m curious: I’m in an online music community that’s been going for more than 15 years (not running on Discourse, BTW). My guess is that the median age of active users is probably late 50s. But many users are much older – in their 70s. Occasionally a user disappears, and then reappears after months and reveals that they had a medical issue, or some other difficult life event. Sometimes we found out (or determine through channels) that a member has died. And there are many formerly active users who have simply disappeared and we don’t know why – or worse: nobody has noticed that they have disappeared.

  • AI could be trained to recognize changes in user behavior (e.g., a user who always posts in the morning suddenly stops).
  • Users could opt-in to a “check-in network,” consenting to be contacted if they go inactive.

Challenges & Considerations

  • Privacy & Consent: Users should be able to opt-in/out and set their own thresholds for outreach.
  • False Positives: Some users may just take breaks; ensure the system doesn’t become intrusive.
  • Data Storage & Security: Tracking user activity patterns could raise privacy concerns.
  • Moderator/Admin Oversight: Should there be a human in the loop (I think so), or is it fully automated?
5 Likes
2

Yup, I imagine the AI Sentiment module could be modified to do something like this. You are right about the challenges and considerations though. I floated an adjacent idea to our internal team and there was a lot of pushback around privacy and surveillance.

6 Likes
3

I love how you care for your members :hugs:

I think this is what will break it in the end :confused:
If you channel all alerts to a single person/group, they will get overwhelmed fast.

Maybe people could have a dedicated “buddy” that gets alerted by the algorithm? That way the burder gets spread out over a larget group of people. And as a bonus, it’s a way to foster interpersonal relationships (I can imagine a lot of messages that will go something like “hey, thank you for checking, I was just on a cruise, it was magical, how are you doing?”).

You will most likely need some out-of-bounds communication channel set up when people opt-in. Because if the worst happens, a PM will likely be ignored by the caretaker. But a direct email might get their attention, a phone call would be even better.

3 Likes
4

I was thinking more about this, and there are probably some models to borrow – particularly from fundraising, subscription, and CRM systems.

The first step might not be a full-on welfare check, but instead would be an automated email to encourage more engagement: “We noticed that you haven’t visited or posted in a while. There’s lots going on. Come back an see whats happening. Share your latest whatever.” And the options (links) would be:

  • Ignore or do nothing (user does not click anything)
  • Visit the Forum Now
  • I’m fine. I’m just busy. Thanks for thinking of me, algorithm.
  • Don’t bug me again (unsubscribe from emails such as this)

And the human “Welfare Check” protocol would kick in only for the “ignore or do nothing” subset after a period of time (one week or…)

5 Likes