Why Do People Accept Policies?

Download Report

Transcript Why Do People Accept Policies?

National Research Council,
Institute of Cognitive Science and Technology
V.LE Marx 15, 00137 Roma
AI, Cognitive & Interaction Modelling
Intelligent Interactive Software
Cognitive Modelling
&
Believable Agents
Laura Benigni
Sergio Benvenuto
Cristiano Castelfranchi*
Amedeo Cesta
Rosaria Conte
Rino Falcone
Multi Agent Systems
&
Social Simulation
Maria Miceli
Angelo Oddi
Mario Paolucci
Roberto Pedone
Giuliano Pistolesi
Paola Rizzo
*University of Siena, Science of Communication
National institue of Statistics
Social & Institutional Influence.
Why people accept policies
Project on
“Multi Agent Systems & Social Simulation”
The Problem
"...to pose a goal to oneself is something about which no external
legislation can interfere...". An agent: "cannot undergo any
obligation other than what he gives himself on his own. (...) only
by this means it is possible to reconcile this obligation (even if it
were an external obligation) with our will".
Kant (Die Metaphysik der Sitten, 1794)
• Social agents are autonomous
• How to regulate autonomous entities?
Organization of the Talk
•
Autonomous agency
– Principle
– Two filters
• Belief filter
• Goal filter
•
Social Influence
– Mechanisms
– Power (e.g.,reliability)
•
•
Goal acceptance
Institutional influence
– Acknowledgement
– Reliability (what is it to trust an institution?)
– Empowerment
•
•
Norm acceptance
Questions
– Motives for acceptance
– Respective efficacy.
Main concepts
• Agents: systems oriented to achieve states in the world.
• Goal: an explicit representation of a world state which the agent wants to be
realised; agents with goals and beliefs are cognitive agents.
• Belief: a representation of the world that the agent holds true.
• Norm: an obligation on a set of agents to accomplish/abstain from a given
action,
– external: no mental representation
– internal,
Social norms are included.
• Institution: a supra-individual system deliberately designed or spontaneously evolved to
regulate agents’ behaviour.
• Autonomy: an agent is autonomous wrt
– its physical environment or
– other agents in the same environment -> social autonomy.
• Goal-autonomy
• Norm-autonomy
Influence
• Principle of autonomous agency
(agj has p as a new goal ) 
q ( agj come to BEL (p  q))
in words, agj has a new goal, if it believes that this is
instrumental to an old one!
• Cognitive influence of agi on agj
(agi GOAL (agj GOAL p)) 
(agi R-GOAL
(agj BEL q p (p  q))
(agi BEL (agj GOAL q)))
in words, if agi wants to influence agj cognitively, agi will let
agi believe that the world state agi wants agj to want,
achieves one of agj‘s goals.
Influencing Mechanisms
• Generate a goal anew (spare power
consumption)
• Modify the value of a current goal (quit
smoking)
Autonomous Agents
• Two sequential filters
– Belief filter
– Goal filter
• An integrated processing of mental representations:
–
–
–
–
Criteria for knowledge acquisition (filtering
beliefs) and belief revision
Mechanisms for goal-activation
Goal acquisition (filtering goals),
Goal dynamics (the change of values and positions
of goals);
Autonomous Agents Architecture
inactive
goals
goalact ivation
beliefs
act ive
goals
goalgenerat ion
(goalfilt er)
Decisionint erruption making
belieffilt er
abandoned
goals
suspended
goals
int erruption
pursued
goals
planner
re-act ivat .
execut able
goals
The Belief Filter
•
To believe is a decision-based state of the mind. Agents have control over their own
beliefs:
– Epistemic control
» credibility :
»coherence with previous beliefs;
• reliability of the source;
• introspection: the agent is the best expert about its own mind
» non-negotiability. The Pascal law: we cannot use threats ("Argumentum ad
baculum") or promises (non-natural means-end links) to make people believe
something.
– Pragmatic control
» relevance : A belief is relevant for a given goal when
• it is about that goal:
x wants p and
x believes he wants p
• it is about the goal's propositional content:
x wants p
x believes that if q is true then p is true also
• it is about the goal's planning links
x wants p
x believes that if p is true then z is true also
x wants z.
The Goal Filter
Agents have control over their own goals:
• Self-interested goal-generation.
An agent is autonomous iff whatever goal q it comes to have, there is
at least a goal p of that agent, for which q is believed by that agent to be
a means.
But self-interestedness ≠ utility function: autonomous agents are not
necessarily aimed at maximizing their utility, they are aimed at achieving
their goals.
• Goal-generation ≠ goal-execution: an agent may abandon its goals.
• Belief-driven goal-processing: any modification of an autonomous agent's
goals can only be allowed by a modification of its beliefs.
Social consequences...
Goal-Generation
• A believed antecedent of a wanted consequence is wanted:
IF x wants p, and
x believes that if q is true p will eventually be true,
THEN x wants q.
( goal≠ executed goal…)
 Beliefs provide reasons for goals...
Goal-Dynamics
… and for changing goals’ values.
•
•
If a belief is retreated, the goal is removed from its current step.
It can be interrupted and put to wait, or be completely abandoned, etc.
–
–
–
–
–
–
–
–
Sleeping (ex. survival: emergency belief -> no move)
Active (ex. now, I want to drink some water: activating perception
Waiting (I give up drinking water now, I’ll do it soon after my talk...:
compatibility belief )
Pursued (... and choose to continue my talk: preference belief)
Planned (ex. How do I keep their attention? Perhaps, I should speed up
a little bit: know-how belief);
Satisfied (I am already speeking at a reasonable speed: satisfaction
belief)
Dropped (perhaps I could tell a nice joke? Impossibility belief
Executed (I promise I will soon get to the end of the talk: Cando belief).
Goal Acceptance
• Goal-acceptance= a special case of goal-generation: social
goal-filter.
–
IF x wants p, and
–
x believes that IF y obtains q
–
THEN x obtains p
–
THEN x wants that y obtain q.
• Autonomous agents accept a new goal iff they believe that it is a means
for an old one.
• The value of a current goal p increases if agents (are led to) believe that
p is
– Instrumental to one more important (meta-)goal q, or more (meta-)goals Q
(instrumentality beliefs. These include beliefs about achievement costs).
– Probability of instrumental connection is higher than expected (probability
beliefs, whose credibility increases as a function of credibility of sources.
These include a different evaluation of feasibility).
– Endangered. Maintenance goals are more compelling than achievement
ones (emergency beliefs).
Influencing Power
• Attractiveness, success, self-confidence, etc.
• Coercive power
• Manipulating power
Etc.
Reliability
– Credibility of BEL supporting GOAL
• Competence of Source
• Sincerity
– Goal importance and risk
Trust = BEL credibility * GOAL value .
(Social) Institutions.
An Operational Notion
• System explicitly designed or spontaneously
evolved to regulate agents’ behaviour in a
common environment.
• Utilities:
– Improve coordination and coperation
– Produce externalities (including conventions)
– Control common/collective resources
• By means of norms and policies.
Institutional Influencing Power
• Acknowledge-ability
• Reliability
• Empowerment (inter alia, to coerce)
Acknowledgement
(From Conte, Castelfranchi, Dignum, 1998)
• Input = a candidate norm (external norm). An obligation in the form
OyX( q),
q = the norm, y = authority that issues the norm and X = the set of the norm subjects.
• Output = possibly a normative belief. Several tests:
– evaluation of the c- norm: is it based on a recognised N?
– evaluation of the source: Is agi entitled to issue N? This entails:
• is q within the domain of y 's competence?
• is the current context the proper context of q?
• is X within the scope of y 's competence?
– evaluation of the motives: is q issued for agi 's personal motives?
The evaluation process is formalised as follows:
– BELx(OzU( r)) & BELx(OzU( r)  OyX( q))
– (OyX( q) & BELx(auth(y,X,q,C)) & BELx(mot(y,OK)))  BELx(OyX( q))
Both lead to BELx(OyX( q))
The relation “auth”: y has authority to issue q on X in C.
The relation “mot”: y's motives are correct.
(10)
(11)
Institutional Reliability
What does it mean to trust institutions ??
• Institutional credit by definition
• Circuits of institutional trust
– Nested :
• Trust a complex agent and therefore
• its members
– Competence
– Disinterestedness (consequent to acknowledgement)
– Tutoriality (“if she is a police-woman, she must be brave…”)
– Transitive
– Select
– Empower
– Control its representatives
 Agents inherit institutional credit but may disconfirm it.
– From institutional trust to goal importance (“if p is of institutional
relevance it must be important…”), and from this back to trust.
Empowerment
(cf. Sergot & Jones, 1995)
• Contributes to
–
–
–
–
Acknowledgement
Trust
Direct influence
Coercion
• Consists of
– Resources (guns, gowns, etc.)
– Action plan and procedures
– Emergent properties of actions and plans (“With the
power conferred by the Law, I hereby declare…”).
Acceptance
(From Conte et al., 1998)
• Is N-belief sufficient? No! Belief about instrumentality.
• Normative corollary of social autonomy: x will form a N-goal q iff it believes
that q is instrumental to a further goal:
BELx(OyX( q)& INSTR(OBTX(q),p) & GOALx(p|r))  N-GOALx(OBTX(q)|GOALx(p|r) & r)
(12)
• Important differences from the g-generation rule:
– the existence of a N-belief.
But norms can be autonomously created:
BELx(O(OyX( q)) & INSTR(OBTX(q),p)
&
GOALx(p|r))
 N-GOALx(OBTX(q)|GOALx(p|r) & r)
(13)
– the form of the instrumental belief. But x may have internalised the norm:
BELx(OyX( q) & INSTR(q,p) & GOALx(p|r))  C-GOALx(q|GOALx(p|r) & r)
(14)
• No N-conformity. We need:
BELx(BELy(OzX( q)))  BELx(OzX( q))
(15)
BELx(N-GOALy(OBTX(q)| r)  INSTR(OBTX(q),be_like(x,y)))
(16)
plus
GOALx(be_like(x,y)|true)
(17)
So far...
• Agents undergo social influence, that is they are often
implicitly or explicitly requested to accept new goals.
• Institutional influence is a special case of social
influence.
• In both cases, autonomous agents accept new goals
(including normative ones) only as means to achieve
old ones.
• Questions
– But what are the specific motives for accepting influence
and forming new goals?
– What is their respective efficacy? Which type of influence is
more effective?
Motives for Acceptance
• Trust (probability/emergency
belief)
Goal (old)
Bel
(p of connection)
Goal
(execute action)
Emotions
• Acknowledgement
Norm
(acceptance)
Bel
(instrumentality)
• Social Responsibility
– Don’t harm
 Material (e.g., passive smoking)
 Symbolic harm (break institutional
authority)
- Don’t give a bad example
Bel
(emergency)
Norm (old)
Goal
(execute action)
Motives for Acceptance (cont’)
Incentives
• Negative
– penalty
– costs of action
– obstacles
• Positive
Bel
(instrumentality)
Bel
(importance)
– side-goals
– meta-goals
Bel
(importance)
Goal
(avoid penalty
Norm
(acceptance)
Goal (old) Side-Goal
Goal
(execute action)
Meta-Goal Goal (old)
Goal
(execute action)
Motives for Acceptance (cont’)
• Social Control
Goal (old)
– Image and reputation
» Responsible
» Rational, consistent
» Trustworthy
Bel
(instrumentality)
– Social isolation
– Social identity
• Sharing (new) social norms
& values
Goal
(accept influence)
Goal
(execute action)
Bel
(value or norm)
Bel
(instrumentality
Norm or Value
(shared)
Goal
(execute action
A Comparison
Criteria (Dawkins, 1976)
• Effectiveness
– Fertility(range of influence)
– Rapidity (how long it takes for fertility to reach
threshold)
– Stability
• Fidelity (no copying-errors)
• Transferability (Cavalli-Sforza & Feldmann, 1985)
– Horizontal
– Oblique
– Vertical
Hypotheses
Stab.
Fert.
Fidel.
Rap.
Tran..
Social t rust
++
+
--
-
+
Inst itut ional t rust
-
-
+
++
++
Side-goals
+
-
-
-
+
Meta-goals
+
-
-
-
+
Avoid sanct ion
-
+
+
+
-
Respons.
++
+
+
-
++
Image/Reputat ion
+
+
-
-
++
Group conformit y
-
+
-
-
++
Avoid Isolation
+
+
-
-
-
Preliminary Conclusions
(to be checked by means of simulation)
 Trade-off between social and institutional influence?
- Social: low fidelity and low rapidity Vs fertility and transferability.
- Institutional: high rapidity and fidelity Vs low fertility and transferability.
 Which one is better? It depends on aims… (e.g., emergency requires
policy-making).
 Institutional influence should be supported by social control. This may
provide guidelines for policy-making. For example,
– Emotions or value-oriented?
– Fear-inducing beliefs or responsibility?
– New values (fitness, etc.) or side- and meta-goals?
 But moreover, which specific type of influence is more effective than
ohers:
– trust-based Vs positive incentives Vs sanctions Vs reputation spread etc.