CAPITULO I MARCO TEÓRICO
1.2. Definición de términos básicos
1.2.2. Aprendizaje colaborativo
1.2.2.3. Aproximación Pedagógica
1.2.2.3.4. Nuevas tecnologías y aprendizaje colaborativo
Once we understand which paradigm to employ with simulation modelling, a lot of criticisms of it from other types of researcher drop away. To those who ask “what if the model misses something out?” we retort that all models are simplifications - they
all miss something out or represent something wrongly (Sterman, 2000, chapters 1, 21). The perfectly accurate representation of human agents with energy would be human agents with energy - but then the real system is not a model, and certainly not one we can experiment with as we please. We can always make our models more complicated, by adding extra factors and extra details to existing factors - but complication rarely enhances communication. The role of the simulation as a tool for
learning and thinking in groups (Morecroft & Sterman, 1994) is undermined by a pursuit of model complexity (Morecroft, 2007, p.412).
Some are suspicious of the simulation model as forming a “black box” - no one understands how it reached its output from its input. Lehtinen & Kuorikoski (2007) identify this as part of the explanation for top economics journals’ relative lack of papers on simulation modelling. They suggest the activity that passes for “understanding” in economics is derivation of theorems from assumptions - if one can perform the logical deductions oneself then one is recognised as having understood, and one feels this is so. Deduction is exactly the part the computer takes over in simulation modelling, so it is no surprise if academic economists find its use a problem.
But being a “black box” is not sufficient reason for avoiding them. Sociologist of science Latour (1987) argues black-box formation is actually part of science. We cannot question every assumption, experimental procedure, or scientist’s competence, and few of us are in a position to question any of them. Over time more and more of these have become unquestioned and disappeared from our attention. In this light the sheer reliability of the computer simulation - its slavish reproduction of the same output given the same program and input - and the ease with which its programs become modules for constructing new programs make it an excellent basis for creating such black boxes. The reliability of the black box allows us to focus on what surrounds it: the assumptions on which it is built; the data it is populated with or validated against; the interpretation of its output, and; the discussions of problems it is
intended to facilitate. From this perspective economists’ love of the deduction would seem to be distracting them from more important business.
COLLINS: Emotional Energy Cultural Capital Interaction Rituals Diffusion of Innovations
RYAN & DECI: Intrinsic Motivation CROSS & PARKER: Energising Social Network Analysis
Baker & Quinn Positive Organisational Studies GROUPS ENERGY SOCIAL INTERACTION CULTURE Durkheim: Solidarity Goffman PROBLEM- SOLVING PERFORMANCE Ethnographic Studies: Communities of Practice Social Capital: Brokerage & Closure AGENT-BASED / SOCIAL
SIMULATION Operational Research:Simulation Modelling Optimisation Axelrod:
Cultural Model
Carnegie School: Computational Organisation Theory
Bounded Rationality Heuristic
Search Algorithms
Figure 11 A coalition of concepts
Illustration of how a simulation model of agents with energy can attempt to bring together concepts of culture, groups, energy, social interaction and problem-solving performance. In this way, a coalition is being attempted that includes three sources of energy concepts - sociologist Collins, social psychologists Ryan and Deci, and social network analysts Cross and Parker - and simulation modeller Axelrod, as well as the literatures on social capital, diffusion of innovations, communities of practice, computational organisation theory, heuristic search and Operational Research.
Latour’s actor-network theorist’s view of science conceives of scientists’ work as the formation of socio-technical coalitions - not just alliances of people, but also materials, machines, theories and ideas. Figure 11 illustrates how this thesis represents an attempt to form a coalition. We have chosen three sources for a concept of energy (Chapter 2) - Collins, Ryan and Deci and Cross and Parker - and we will try to encode a concept that unifies all three. Collins, we have noted, connects us to
concepts of culture, groups and social interaction rituals. But a theory that unites these may explain some of the problems in the diffusion of innovations, as identified in the literature on both social capital and communities of practice (see section 1.3). Axelrod’s cultural model is a social simulation that already links culture, groups and diffusion, but does not include energy or problem-solving performance (see section 3.4.2). Interest in problem-solving in organisations was one of the key features of the Carnegie School (section 1.2), as was their stress on bounded rationality and heuristic search. Operational Research contains expertise in simulation modelling and heuristic search algorithms (which we draw upon in Chapter 6, in ways reflected on in Appendix F), and has applied its tools and techniques for problem solving in organisations. No one else has attempted to fill the “structural holes” (Burt, 1992) between all these areas - Baker and Quinn’s (2007) energy model omits culture and groups for instance (see Chapter 4), and all that they link to - so success would constitute a contribution to knowledge. Equally, identifying previously unforeseen problems in relating these areas would also be a contribution.
Armed with this network we can see how we connect to empirical work: Cross and Parker’s social network data collection and interviews; Ryan and Deci’s laboratory experiments; Collins’s theoretical sociology based on various secondary data sources; ethnographic studies of communities of practice. There are also various studies on social capital and on diffusion of innovations. The empirical grounding may be indirect, but it is there, and it should raise confidence in the value of our project.
Raising confidence in the execution are our links to the standards of simulation modelling in Operational Research, and to other modelling experience - especially that using the Axelrod Cultural Model (see sections 6.2, 6.3 and 8.2).
Future tests of this research would include attempts to reconstruct our models and replicate our results, as well as attempts to produce models by other means for similar ends, and extensions of the work. Such replication is thought “one of the hallmarks of cumulative science” (Axelrod, 2006), but Latour (1987) questions whether it is particularly common now in natural sciences where the expertise, facilities and time are rarely affordable. In the literature on agent-based social simulation actual attempts are certainly rare, and not without their problems (Axtell et al, 1996; Axelrod, 2006). In our own experience, attempts at reproduction led to us identifying a minor bug in a sampling process in Axelrod’s Cultural Model (Axelrod, 1997a, chapter 7; 1997b), and an erroneous description of the calculation of performance in March’s much-cited model of organisation learning (March, 1991) - made worse by a chart that started its scale at 0, despite the model producing some negative output values. An attempt by several experienced modellers to replicate eight published agent-based models “identified problems with respect to ambiguity, gaps, and even errors in the published descriptions, as well as subtle differences between how different floating point systems calculated…” (Axelrod, 2006). They found three decreasing levels of replication:
“‘numerical identity’ in which results are reproduced precisely, ‘distributional equivalence’ in which the results cannot be distinguished statistically, and
‘relational equivalence’ in which qualitative relationships among the variables are reproduced.” (Axelrod, 2006)
Practice in replication also builds capability for constructing one’s own models in the future, and extending the work of others. Axelrod’s prescription for progress in social science simulation is “methodology, standardisation, education and institution building”. Our experiences would add the importance of making code available to others - on the web as Axelrod did for his cultural model (Axelrod, 1996a) or on request as March did to one researcher who helpfully then reproduced March’s formulae in his extensions (Rodan, 2005). Re-engineering a model from its author’s description is a good verification exercise, and producing one’s own model to serve a purpose addressed by another’s can reveal the value of both. But when the results fall short of numerical identity, comparing definitions and algorithms in the code is instructive and relatively quick once one has some idea of what one is looking for.
Finally there is the question of how we feedback into the literatures we draw upon. As we noted above in 5.3 simulation modelling as coalition formation can identify inconsistencies and ambiguities in an author’s presentation of their ideas, tensions between authors, and unanticipated and unwanted logical implications. In stochastic simulations it can trace the distributions of outputs, and in complex models it can reveal interdependencies between variables and sensitivities to initial conditions. Our findings may be presented to the authors we drew upon, and to others working in the same areas. In addition, we can return to the literature with fresh questions in mind, and seek out further works by these authors. A degree of validation of this project comes through finding that previously unseen material fails to introduce more
tensions into our coalition, though current absence of tensions is no absolute guarantee of future absence. Confidence in a model or in a coalition can be raised, but not to an absolute value.
If simulation is a means of theory development (section 5.3), what denotes good theory? In a much-cited paper Weick (1989) argues for a focus on plausibility rather than validation, and the raising of possibilities, not predictions. When we present to another person, four reactions substitute for validity, the social scientist’s equivalent of significance tests:
• “That’s interesting” (a moderately strong assumption is disconfirmed);
• “That’s absurd” (strong assumption disconfirmed);
• “That’s irrelevant” (no assumptions activated);
• “That’s obvious” (strong assumption confirmed).
Other reactions include:
• “That’s connected” - commenting on relations to other thoughts;
• “That’s believable”;
• “That’s beautiful” - though aesthetic reactions are more common in mathematics;
• “That’s real”.
This latter reaction can be the occasion of performing a reality check which protects against shifts:
• from “that’s interesting” to “that’s in my best interest”;
• from “that’s obvious” to “that’s what managers want”, or;
• from “that’s believable” to “that’s what managers want to hear”.
But “that’s real” can also represent a shift itself to “that’s the power system I want”.
Weick sees theory construction as “disciplined imagination”, a process likened to evolution by natural selection. Through use of the imagination we create alternative versions of existing theory, thereby introducing variation. Discipline - including tests of consistency and plausibility - provides selective pressures so that only the fittest survive.
Simulation aids this process in both respects. As the philosopher Dennett has put it, “What you can imagine depends on what you know.” (Pyke, n.d.) Indeed, elsewhere Dennett has described how John Conway’s cellular automata, the Game of Life, acts as a “tool for thinking about determinism”, since experiencing it expands one’s imagination for thinking about this philosophical issue (Dennett, 2004). Closer to our purpose here, simulation’s power for exploration and experimentation helps us generate new conjectures for testing. The rigour and precision of coding, together with the demonstrations of logical implications, discipline us in our creation. While the modelling process, as a tool for facilitating dialogue, adds the selective pressures provided by participants - who may include those imbued with the values of quantitative or qualitative research.