Beehive moderation

There is a simple idea for decentralized moderation without hierarchies. You allow anyone to remove any other member like is allowed in Delta Chat already, but the member who removes someone has to leave as well.

This disincentivizes removing random members. If the member removes actual spammer or troll, the group will add such member back in later. If a troll removes random member without a reason, the group will add member back, but not the troll.

As far as I know the idea originates in the paper Suicide for the Common Good: a New Strategy for Credential Revocation in Self-Organizing Systems (2006). You can read the PDF at https://cs.wellesley.edu/~tmoore/OSR-suicide.pdf

There are however some problems for practical implementation of this:

  1. An attacker can add as many puppet profiles as already in the group first,
    then completely destroy the group by removing everyone.
    This is referred in the paper as Sybil and replication attacks.
    A solution may be to disallow new members adding members
    until they are in a group for 30 days,
    some voucher/ticket system or voting
    (which may be better than just introducing voting for removals
    because it allows to quickly remove members in case of abuse).
  2. The paper implicitly assumes the existence of broadcast channel.
    An attacker may send a removing Alice
    to half of the group and a message removing Bob to half of the group.
    Then the group will have to figure out whether Alice or Bob should be removed,
    ideally invalidating both removals and only removing the attacker.

This idea had some popularity in MANET research and there are many related papers continuing this theme at Google Scholar, for example: The fable of the bees: incentivizing robust revocation decision making in ad hoc networks (2009) (The fable of the bees: Incentivizing robust revocation decision making in ad hoc networks for CCS 2009 - IBM Research) with a PDF available at https://web.archive.org/web/20170809071452id_/http://amnesia.gtisc.gatech.edu/~moyix/CCS_09/docs/p291.pdf

Related topics

1 Like

I want to try to respond without claiming to know enough about the subject.

I take it that the group is a convention shared among all the clients that are members of the group; a participant’s client accepts incoming messages from the group and relays them to the others.
So there is no real subscription to the group but only the participation of a user and his presence in the list of others.
So I infer that being kicked out of a group corresponds to the removal of one’s address from the broadcast list of others. The group is not a thing, it is a room without walls

Then the idea of committing suicide for the sake of the common good is not a bad idea but in reality this does not translate as being thrown out of a group but as the refusal of other clients to pass on messages to us anymore.

So in my opinion the problem is not centered on who you decide to throw out but who you decide to let in.
I would therefore try to solve it in this way:

  1. a person can join a group when the majority of the participants in a recent period (I don’t know how much, say the last three days) agree to let him in. This prevents someone from letting in puppets.
  2. if the same amount of people at a given time (the majority relative to the recent time) agree to kick a person out, then this choice will propagate to the remaining clients.

There should be well calibrated limits, how much time to consider counting active users, how long it takes for a decision to be made to kick a person out or not and the minimum number of people who must participate in the vote.

Then if even in this way there are more trolls in a group than participants, we have to come to terms with the fact that the group was not created with proper care and has the right to explode. After all, if most of the participants are lesionists, why would the few healthy participants decide to stay in this group and not create another one?

I really like that there are no hierarchies in deltachat, and a bit of an example of implementing a utopia where people are mature enough to self-manage without having to necessarily submit to each other. If it does not work, it is the fault of human nature.