From Santa Fe Institute Events Wiki

Revision as of 06:26, 25 June 2010 by Mlaidre (talk | contribs) (→‎Gossip Project)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Gossip Project


Gossiping about complex systems: a network model of gossip assessment

By The Santa Fe Gossip Group

Gossip—talking about third parties behind their back—is a uniquely human trait, one not found in any other species. We humans also spend a significant amount of our lives gossiping, with roughly one to two thirds of our conversation time focusing on gossip. While gossip can have positive functions (enabling listeners to keep track of others without first-hand observation) gossip is also problematic since it is susceptible to corruption. A key theoretical question is therefore how individuals can sort true from false gossip, using the true gossip to accurately reconstruct the going-ons within their community. Here we present a simulation model in which gossip propagates through a network of agents, some of whom distort information by propagating bogus gossip. We investigate the fitness of different decision-making rules for assessing the veracity of gossip and we provide preliminary results for how well these rules perform under different societal starting conditions. In the vein of Axelrod’s computer tournament on cooperation and Laland’s computer tournament on social learning we propose to undertake a computer tournament on gossip assessment. We hope you will submit a program. And may best gossiper win!

Next project meeting 6: 30 Thursday after dinner

Tuesday dinner meeting: Mark Laidre, Anna Packenkina, Megan Olsen, Daniel Jones, Erika, Kaisa, Dan MacKinlay, Susanne Shultz

Suggestions (SS): Could we have a stub with recommended papers either listed or uploaded? The following are my notes from the meeting, please comment/edit

Project concept

Evolutionary dynamics models of gossip networks and honesty:
  1. Social rules/norms and norms
    1. Use sociology gossip models
      1. Reliability may be a function of number of informants (the more confirmation of a piece of gossip, the more reliable the information)
      2. Help from sociology literature????
    2. How does bogus information/informants (false/distorted gossip) affect optimal rules
    3. Extensions: influence of positive versus negative information, in group versus outgroup impacts
  2. Future possibilities?
    1. How does network structure impact on gossip dissemination?
    2. Parameterise models based on realistic networks
    3. How does gossip impact on stability of cooperation?
    4. Gossip and the evolution of language?


  1. Structure
    1. Graph versus agent-based models
    2. If there is no movement of ‘agents’, then it is probably better to use graphs.
    3. Network based model (use realistic parameters)
    4. Start with netlogo?
  2. Parameters
    1. Stable network
    2. Subgroup that has information about a third party
    3. Can propagate (at different levels)
    4. Can have reliable information
    5. Extensions/additional parameters: Secondary rumours, memory, trust, Hebbian learning (mentioned but not sure in what context)???
  3. Fitness functions (per interaction/click?- absolutely critical to define these, since they are core of model)
    1. Target
    2. Honest signals
    3. Bogus gossipers
    4. Listeners

Related issues/models

  1. Byzantine generals and traitors
    1. lamport et all's classic paper gives rigorous results about defector detection
    2. Google's take on the complexity of their implementation (i.e. people probably don't execute "thousands of lines of C++ code" when deciding who to believe)
  2. Empirical tests?
    1. Mafia Game (more information please)
    2. Try to seed gossip statement and see how they travel through the network based on their 'importance'