Upcoming SlideShare
×

# Norms recognition

389 views

Published on

Published in: Technology
0 Likes
Statistics
Notes
• Full Name
Comment goes here.

Are you sure you want to Yes No
• Be the first to comment

• Be the first to like this

Views
Total views
389
On SlideShare
0
From Embeds
0
Number of Embeds
6
Actions
Shares
0
7
0
Likes
0
Embeds 0
No embeds

No notes for slide
• The moduel includes a ltm and a wm. In the ltm, norms are archived once nbels got formed. The wm is a 2layers architecture, where new inputs are elaborated. So anytime a new input vector arrives, if it is a D or V, it will access the wm at the 2 layer, and a N-bel is formed. If not, but it has already been ancountered as a D or V, it will access the wm at the first florr. If not, it is matched with the NB and if…, otherwise it is discarded. This architectire allows us to treat the gradual formation of beliefs. If , within a certain time periodo, the CNB is not strengthened by new inputs (N-threshold), it will exit the wm, otherwise a fullfledged Nbel is formed and archived.
• ### Norms recognition

1. 1. N-Recognition Module Input (Candidate N-b el)‏ N-bel V c =N-threshold E N-Board LTM B  , R  , A  > v c D  , V  < v c W M
2. 2. A simulation study
3. 3. Norm-recognizers Vs Social conformers <ul><li>What are observable effects of norm recognition? </li></ul><ul><li>Implement different populations (Andrighetto et al., 2008a, Campennì et al., 2008b): </li></ul><ul><ul><li>Social conformers: follow actions most frequently done in observation window (parameter)‏ </li></ul></ul><ul><ul><li>Norm recognizers take input from others, form beliefs and act based on those. </li></ul></ul>
4. 4. Agent and world <ul><li>4 contexts: </li></ul><ul><li>following its agenda and time of permanence, each agent moves among contexts; </li></ul><ul><li>in each context, agents can produce 1 out of 3 actions; </li></ul><ul><li>1 action is the same for all of the contexts; </li></ul>C1 (A1, A2, A3)‏ C2 (A1, A4, A5)‏ C3 (A1, A6, A7)‏ C4 (A1, A8, A9)‏ AGENDA (C1,C2,C4,C3)‏ Time-of-Perm (nTicks/ n )‏
5. 5. Norm recognizer <ul><li>Each is provided with: </li></ul><ul><ul><li>Normative Board </li></ul></ul><ul><ul><li>Double-layer architecture </li></ul></ul><ul><ul><li>Agenda: individual time of permanence (in contexts). </li></ul></ul><ul><ul><li>New normative beliefs contribute to choose action </li></ul></ul><ul><ul><li>if normative board is empty, action is randomly chosen. </li></ul></ul>N-Board: N-B1 N-B2 ....... level-2 (D & N-V)‏ level-1 (A, R, B)‏ AGENDA (C1,C4,C2,C3)‏ Time-of-Perm (t)‏ N-Threshold (v c )‏
6. 6. Social conformer <ul><li>Each observes other agents in same context </li></ul><ul><li>According to conformity rate, imitates most frequent action </li></ul>Conformity rate = 9
7. 7. Simulations' results
8. 8. Preliminary Findings <ul><li>Social conformers (above): </li></ul><ul><ul><li>Each colour represents one action </li></ul></ul><ul><ul><li>No difference within ticks </li></ul></ul><ul><ul><li>Strong difference </li></ul></ul><ul><ul><ul><li>Among ticks (no belief)‏ </li></ul></ul></ul><ul><ul><ul><li>Among scenarios (no memory)‏ </li></ul></ul></ul><ul><ul><ul><li>More frequent action (dark blue) is distributed throughout the simulation: nothing emerges! </li></ul></ul></ul><ul><li>Norm recognizers (below): </li></ul><ul><ul><li>Fuzzier </li></ul></ul><ul><ul><ul><li>Rows (autonomy)‏ </li></ul></ul></ul><ul><ul><ul><li>Columns (beliefs)‏ </li></ul></ul></ul><ul><ul><li>After 60th tick, one action common to all scenarios: something emerges… </li></ul></ul><ul><ul><li>What is it? Lets look into agents beliefs… </li></ul></ul>Social conformers Norm recognizers
9. 9. Immergence <ul><li>At the 30th tick a normative belief starts to spread </li></ul><ul><li>What has happened in the in the interval? </li></ul><ul><li>Other normative beliefs got formed, although earlier is more frequent </li></ul><ul><li>Immergence is earlier: it takes time for effect to emerge </li></ul>
10. 10. SCs Vs NRs Low Conf. Rate 3 Medium CR High CR Medium NT Lo N-Threshold Hi NT
11. 11. Latency of norms <ul><li>Time interval between N-bels appearance and convergence on corresponding action. </li></ul><ul><li>Actually, a complex loop </li></ul><ul><ul><li>from N-Bel x to N-action x </li></ul></ul><ul><ul><li>from N-action x to N-bel y </li></ul></ul><ul><ul><li>from N-bel y to N-action y </li></ul></ul><ul><ul><li>Etc. </li></ul></ul><ul><li>Can help predict effect of new policies </li></ul><ul><li>Immergence  2nd-order emergence: not a reflected upon emergent phenomenon but involved in emergence! </li></ul>
12. 12. Follow-up questions <ul><li>Only common actions? </li></ul><ul><li>What happens with physical barriers and/or (cultural) drifts ? </li></ul><ul><li>Equally frequent norms might emerge in different sub-populations: norm innovation? </li></ul>
13. 13. Let us simulate a barrier <ul><li>At a given tick, agents get stuck to current locations, they can no longer move across settings. </li></ul>Yes barrier No barrier
14. 14. Normative Beliefs No barrier Yes barrier
15. 15. But in 300 ticks…
16. 16. Final remarks <ul><li>In a multi-scenario world, unlike social conformers, norm recognizers converge. </li></ul><ul><li>Norms immerge in the minds before emerging in behavior. </li></ul><ul><li>A normative belief corresponds not necessarily to the most frequent action. </li></ul><ul><li>Barriers are sufficient (not necessary) for norm-innovation. </li></ul><ul><li>Norms have a latency time. </li></ul>
17. 17. For the future <ul><li>If agents get re-united, which norm is going to invade population? </li></ul><ul><li>What about inertia (i.e. the time for a norm to disappear)? During inertia, norms may compete in the same population... </li></ul><ul><li>Internalization and semi-automatic conformity </li></ul><ul><li>Questions for future studies :-) </li></ul>
18. 18. Why bother? <ul><li>Policy-making and what if analysis </li></ul><ul><li>ICT systems: </li></ul><ul><ul><li>Artificial normative reasoning </li></ul></ul><ul><ul><li>Electronic institutions </li></ul></ul><ul><ul><li>Applications of agent systems </li></ul></ul><ul><li>Computational and simulation-based modelling for (social/cognitive) theory-building and testing. </li></ul><ul><li>Thank you </li></ul>