Signaling Games on Networks
• Extension in time & space
• Agents are placed on a social network
• Agents communicate repeatedly with their neighbors
• Agents‘ decisions depend on learning dynamics
Topic 1 Topic 2 Message 2
Message 1
S1 Topic 1
Topic 2 Message 2
Message 1 S2
Invasion & Social Status
• Project: “Language Contact & Change“ (Jason Quinley & Roland Mühlenbernd)
• Research Question: What kind of environmental circumstances explain the emergence of a specific meaning space division in the English language
• Publications:
– Quinley & Mühlenbernd (2012): Conquest, Contact, and Convention:
Simulating the Norman Invasion’s Impact on Linguistic Usage, BRIMS 2012 proceedings, pp. 113-118.
– Mühlenbernd & Quinley (2013): Signaling & Simulations in Sociolinguistics, Pennsylvania Working Papers in Linguistics
Simulating Language Contact
• Salient systematic division of meaning space between words of Germanic and French origin
• Social conditions like the lifestyle of the Anglo-Saxon laborers and the French- speaking nobility causes this divergence
Germanic French
work (Werk) play (plaisir) stone (Stein) jewel (joel) swine (Schwein) pork (porc)
animal meat
"pork"
"swine" animal
meat
"beef"
"cow" animal
meat
"mutton"
"sheep"
animal meat
“fish" animal
meat
"Schwein"
The Context Signaling Game
1. Nature chooses topic (T = {meat, animal})
2. Sender transmits message to Receiver (M = {swine, pork}) 3. Receiver chooses interpretation (A = {meat, animal})
4. Sender and Receiver in context (C= {farm, dinner}) 5. PC(c) depends on sender‘s social status Si
6. Context Influences Topics Pr(t| c)
Dinner Farm
Computer simulation
animal meat
"swine" animal
meat
"swine"
animal meat
"pork"
animal meat
"pork"
animal meat
"pork"
"swine" animal
meat
"pork"
"swine"
animal meat
"pork"
"swine"
SA
SN
SE
SU
Result 1: The whole society learns either SE or SU, but both strategies are equiprobable.
Random Replacement
Trial 1: Replace Agents at Random Why?
Are Conventions Pre-Determined?
Result:
• All Learn Convention BUT
• Strategies Equiprobable
• Contact NOT Enough!
Meat or Animal?
swine pork
beef cow
fish fish
Contact Meets Context
Trial 2: Replace Big Nodes with Invaders
Why?
High Status
- Higher Probability of C
d- Higher Probability of T
mResult: Convergence to Convention
Factoring in Social Status
animal meat
"pork"
"swine" animal
meat
"pork"
"swine"
SE SU
Status S1 S2 S3 S4 S5 S6 S7 S8 S9
Pr(Cd) .1 .2 .3 .4 .5 .6 .7 .8 .9
0 0.2 0.4 0.6 0.8 1
≥ S1 ≥ S2 ≥ S3 ≥ S4 ≥ S5 ≥ S6 ≥ S7 ≥ S8 ≥ S9
Of Nodes and Nobles
Trial 3:
• Replace Hubs with Invaders
• Tie Degree to Status Why?
Preferential Attachment Scale-Free Social Networks
• Power Law Distribution - Degree
- Wealth
Result:
• Faster Convergence
• Context + Connectivity
Future Research
• Game Dynamics
– Cost vs. Capacity
– Punishment/ Prestige – Norms vs. Conventions – Innovation/ Extinction
• Network Dynamics
– Rewiring
– Bottom-Up Invasion (e.g.
Czech)
Standard Colloquial
závin štrůdl
láhev flaška
kniha bichle
German loans in Czech