Barcelona, November 3-5, 2014

Web del Congreso: http://ecsi.sintelnet.eu/
Proceedings: http://ceur-ws.org/Vol-1283/
Organizado por European Network for Social Intelligence (SINTELNET)

Social intelligence is a general term at the intersection between different disciplines including philosophy, social science - sociology, economics, legal science, etc. - and computer science. Broadly speaking, social intelligence is the capacity to understand others and to act rationally and emotionally in relations with others. This is an ability that not only human but also artificial agents have, as modelled in artificial intelligence and agent-based research in particular.

The interactions between philosophy, social sciences and computer science around social intelligence are manifold, and many concepts and theories from social science have found their way into artificial intelligence and agent-based research. In the latter, coordination and cooperation between largely independent, autonomous computational entities are modelled. Conversely, logical and computational models and their implementations have been used in the social sciences to help improve simulations, hypotheses and theories. Among the most prominent subjects at the interface are action and agency, communicative interaction, group attitudes, socio-technical epistemology and social coordination. In computer science, these concepts from social science are sometimes deployed at a more metaphorical level rather than in the form of rigorous implementations of the “genuine” concepts and their corresponding theories. Equally, the computer models used in social science are not always convincing.

The aim of the European Network for Social Intelligence (SINTELNET, 2011-2014) is to help build a shared perspective at the intersection of the above fields, to identify challenges and opportunities for cross-disciplinary collaboration, to provide guidelines for research and policy-making and to kindle partnerships among participants. The aim of the European Conference on Social Intelligence is to provide a productive meeting ground for researchers from the above fields. Topics of interest include but are not limited to:

• Action and agency
• Mental attitudes (e.g., beliefs, desires, goals, intentions)
• Dynamics of mental attitudes (at individual and collective level)
• Argumentation theory
• Cooperation and coordination
• Social epistemology and ontology
• Epistemic game theory
• Theory of signalling
• Norms and normative systems
• Group attitudes, collective reasoning, collective intentionality
• Theories of artefacts and technical design
• Socio-technical systems
• Trust and reputation
• Speech act theory
• Epistemic and evolutionary aspects of conventions
• Emotions and affective phenomena
• Social influence
• Social power and social dependence

http://www.ia.urjc.es/~dpearce/

http://personal.lse.ac.uk/LIST/

Dependencia del contexto por parte del agente.

Deviations from classical rationality: Bounded (framing effects) vs sophisticated rationality (politeness)

Distinguish between the context and the array of options from this context.

Properties: (x,K) option-context pairs Property System: all the options an agent has available

REASON-BASED MODEL explain an agent choice function pair of

• motivation salience function
• fundamental preference relation

Contexts with the same context properties induce the same motivationally salient properties This model explains

Ability to predict future choices

CAPTURE THE COGNITIVE PROCESSES BEHIND THOSE DECISIONS

El estado de animo del agente puede caracterizarse como una opcion mas del contexto

Este modelo se propone con un esqueleto formal (formal shell) que puede ser rellenado con elementos procedentes de la sociologia y la psicologia

Mel Slater, ICREA Universidad de Barcelona

VR: Respond to events and situations in immersive virtual reality. Sirve para estudiar el uso del cuerpo y su papel en la percepcion porque no estas moviendo un raton con los dedos sino que como la experiencia es inmersiva mueves todo el cuerpo porque estas dentro de la situacion.

Hay ciertos aspectos de la mente/cerebro que no conocen la diferencia entre entornos simulados y reales y actuan como si fuera real aunque el individuo sepa que es simulado. Entornos de VR se usan para estudiar la respuesta ante la violencia, algo imposible con experimentos reales (ej. violencia en el futbol o estudios clasicos eticamente problematicos electroshock). VR es eficaz para estudiar dilemas morales aunque la representacion visual no sea realista.

Seminal Milgram paper. El problema del experimento de Milgram es que hacía pensar a la gente que era real y por tanto el experimento incluía más factores de los que se pretendían estudiar. En VR todo el mundo sabe que el experimento es simulado y aún así actúan COMO SI fuera real. Este hecho elimina aspectos morales complejos del que se somete al experimento y se estudian solo aquellos aspectos más intrínsecos.

Body Ownership: induced by multisensory contradictory data

• The Pinoccio Illusion (Lackner 1998)
• The Rubber Hand Illusion. Visual-Tactile Synchronisation (Botvinick & Cohen 1998,Nature) prioperceptive drift
• Illusory Agency. Ver a un cuerpo que asumes tuyo porque se mueve en sincronizacion con el tuyo hacer algo por su cuenta hace que pienses que efectivamente tú lo has hecho. Visiomotor sync “Este es mi cuerpo y habló por tanto yo hablé”.

EMBODIMENT IN A ROBOT, BEAMING (Star Trek), se puede hacer mediante VR. Poseer el cuerpo de un robot en un lugar distante hace pensar al que lo posee que efectivamente está en ese lugar distante.

Peliculas: Código fuente, Avatar

La tesis de Andy Clark de la mente extendida se relaciona con esta capacidad de la mente de incorporar elementos externos al cuerpo como si fueran propios y a usarlos en los procesos cognitivos.

Body ownership

• can be achieved through multisensory illusion
• has consequences depending on the form of the body
• contributes to agency

THESIS: the body has a semantics

The issue of AGENCY crucial to understand as we become replaced more and more by avatars and robots.

Computer games related to violence? literature very divided. Falta investigación sobre la posible relación entre videojuegos inmersivos y violencia. Con la aparición de Oculus Rift se va a facilitar este estudio.

Socio-tecchnial systems require new skills, conventions, a new view on almost everything. Physical and virtual intermixed. Requires augmented body and augmented mind because we live in an augmented reality living at the same time in two worlds.

This organisation cannot be planned, it is an espontaneous order, it emerges.

• Not only bounded rationality (Simon)
• but COMPLEXITY
• but COMPUTATIONAL INTELLIGENCES
• for the intrinsic blindness typical of organized institutions

We need a new Simon for explaining rationality at the collectivelevel

1. GENERAL PERSPECTIVE

The COGNITIVE MEDIATORS of Social phenomena, richer cognitive models for “artificial intelligences”

COGNITIVIZING: cooperation, conflict, power, social values, commitmentnorms rights, social order, trust

Pareto, Garfinkel: social sciences as opposed top psichology. We need to go back.

We need MIND READING because agents behaviours are due to the mental mechanisnm creating and controlling them.

Una teoria del cerebro que evita la mente no permite entender las inteligencias artificiales.

Social interactions are artifacts not only for coordination but to predict and prescribe the mental states of participants. THE CENTRAL DEVICE IS MIND PRESUPOSING AND MODIFICATION.

• We need MIND MODIFICATION models: goal adoption and goal induction, m mind and other's mind
• social coordination works “as if” they have a mind
• MIND is a social artifact. our social minds are social institutions
• ASCRIBED and ENDOWED minds are crucial coordination artifacts because they crete the common ground, shared knowledge.
• COMMUNICATION is also for shapingmind

BUT MIND IS NOT ENOUGH

• The social actors do NOT understand, negotiate and plan
• Identify the MENTAL MEDIATOR. unavoidable alienation, Leviatha Demo-crazy
• Necesitamos entender como construimos algo que no entendemos aún.

MIND NOT ENOUGH - SELF ORGANIZATION

• emergence & inmergence
• emergence cognitive, dependence in network, interference in the world
• spontaneous social order: Friedrich Hayek: emergence must be functional. (Hayek: Knowledge. Market. Planning)
• Adam Smith invisible hand: teleological nature of invisible hand to pursue social order. Ideologism, too much positive. Must be rejected but social order is emergent as Smith said.
• How is posible that we pursue something that is not an intention of ours?

2. THEORY OF FUNCTION

theory of eemerging functions among cognitive agents NEEDED

In an hybrid world we can reduce guman affective handicap providing more reliable data

Social functions require an aextracognitive emergence working the efectiveness of social function is independent of agents understanding of this function on their own behaviour

Two finalistic systems

• goal oriented
• goal governed

Functional OK, teleological no.

KAKO-FUNCTIONS POSSIBLE?

• cannot be explained in behaviouristic or reincorcement scenarios
• notion of function as SELECTING and REPRODUCINGits own causes
• we need COMPLEX REINFORCEMENT LEARNING FORMS operating on GOALS and BELIEFS, thts is, in the cognitive representations
• example of kakofunction: dirty and clean screens
• institutional level vicious circles: prisons reproduce delinquency
• FUNCTION is something SELF REPRODUCING AND SELF PRODUCED, emergent

3. BLIND SOCIALITY

Obey norms blindly make norms work because the issuer see norm as a tool for a problem. We trust that norm is for social good. Socrates taking the poison. But there is a part of the norm that has to be understood partially.

We blindly reify, objectify power. We dress theking with our eyes.

The “mistakes”, like the idea of god, works very well socially. Doesn't depend on existence.

Social Control

• MANTENER CONTROL: delegar en IA el COMO conseguir una meta pero no dejar que escoja CUAL meta conseguir
• OPEN DELEGATION, transparently let know all goals
• AVOID UNAWARE COOPERATION, better goal adoption instead of goal delegation

We need adjustable autonomy

• MONITOR PEOPLE to understand why they need to violate norms: possible danger of formalization and enforcement of rules.
• Violations sometimes produce better functionality.

CONCLUDING REMARKS

• we are engineering a new society
• reconcile emergence, self-organizing with intelligente, people participation
• self organization = out of mind

  • society works thanks to our PARTIAL INTELLIGENCE, not knowing whats going on at social level
  • will invisible hand become a computational invisible intelligence orchestrating societies? PRESERVE SELF-ORGANIZATION
  • reconcile emergence and congnition
  • en sociedades híbridas se necesita información para que la gente que conozca el cuadro completo de normas y todos los efectos
  • alienation
  • worry: net-demagogy
  • Mark Twain: si votar pudiera cambiar el orden social no nos dejarian hacerlo

  The book Computational intelligent data analysis for sustainable development: shows how predicting without understanding is possible in this area also

Science will be computational or will not be

• AI (Artificial Inleggigence) was the first attempt
• not just models but EXPERIMENTAl PLATFORMS, VR

The Goal-Oriented Agents Lab (GOAL) is an interdisciplinary group that carry out research on finalistic behavior in intelligent agents. Key areas of activity are Cognitive Systems, Social Cognition, Action Control, Decision Making, and Emotions. Since the 70s, members of the group developed a novel approach to cognition, known as goal theory. www.istc.cnr.it/group/goal

Q&A

• Formalizacion no es necesariamente mala, lo malo es crear un modelo social en el que la violacion de la norma no este contemplado
• Big data es como la gravedad, desde Newton sabemos como funcona en la practica, es una ley, pero no es una teoria porque nosabemos que es realmente. con Big Data encontramos resultados espectaculares de prediccion minando grandes cantidades de datos pero no entendemos los mecanismos sociales que intervienen.
• social simulations
  • with insects we predict social complexity
  • we eplain without cognitiva agent.thats true
  • technology for collective intelligency
  • VDI is just an preliminary step
  • do we need an emotional mind alwaysin simullations? castelfranchi thiks not
• there is no technical perfect solutions to political problems because the cause is here are CONFLICTING PERSONAL INTERESTS