An Individual-Based Model of Innovation Diffusion Mixing Social Value and Individual Benefit Defuant Huet Amblard 2005

This was the example of a recent paper that shows that this work hasn't changed in 30 years (1978). And i can see that, there isn't really any falsifiability, at least not the way they structured it. It does have lots of parameters though. The idea is to model popularity and fads. They explore the idea that a person is more or less persuasive when they are talking to someone they agree with or disagree with. Modeling this, they get some complicated results that make predictions for society. or atleast 'guide decision making' i thought it was interesting enough, though I didn't really invest in it enough to really consider the possibility offered by this paper of deeper insights into how groups work.

One thing that this paper Did get me thinking about: The results of a model can be predicted in a real phenomenon to the extent that one accepts not only the models assumptions in the reality but the model's omissions as well. My model of economic growth assumes rationality and role of GDP and all that, and it omits social factors, history and persistent inefficiencies. I have to accept the assumptions and the omissions. That was my old account of the validity of experiments on models as replacements for experiments in the corresponding reality. But this paper is going to require me to nuance that.

We read papers that were much simpler than this that, coarsely, had analogous results. Now this model is trying to explain more and it has more parameters. It 'quantifies' things like 'the media', 'social opinion', 'social opinion uncertainty' and 'un/popularity'. Should I say that it is making more assumptions or fewer omissions? I guess both, as long as I am ok with the remaining omissions and all these new assumptions. In a way, it is easier to disagree with; it is a more honest model, because it is more possible to make assumptions explicit than omissions. It is also easier to disagree with because it introduces way too much complexity.

This paper does allow them to make this argument: 'No more than these parameters are sufficient conditions for innovation diffusion in society to the extent that one accepts the models omissions and assumptions.'

Not a very strong case, i guess, but science is incremental (bad consolation).

Notes

• it cites the Granovetter, positioning itself on nuancing his threshold model of 30 years ago. it is interesting that, though it is a forefather, there is no use of any game theory vocabulary here ('actors' vs 'agents', 'optimal', 'rational'. only benefit, but it is used here in a distinctly more lay sense)
• three wyas that it develops/nuances the previous work as a model for social info diffusion:

1. 1. old models assume that people know their individual benefit from an action

• • • (interesting socialogical result cited: people who are itnerested gather more information, people who aren't don't. i know that sounds trivial. maybe it was neat in context)

1. 1. contagion is stronger or weaker for individuals
   2. decisions are not binary

Results (interesting that they don't say 'Predictions':

1. innovation with high social value and low individual benefit propagates better than an innovation with a low social value and high individual benefit
2. A minority of extremists has a large role on adoption when the density of the social network is high and also the frequency of discussion
3. low levels of adoption can be due to high uncertainty abou tthe innovation

Goow quote accomodating almost everything I've said: "

We are conscious that many of the assumptions behind the model are highly debatable, the chose of the social dynamics, the dynamics of discussion, the strong separation between social and personal opinions, the type of social networks, and more.  We made several of these choices with poor or without empirical justification, and sociologists or psychologists might see them as strong and artificial simplifications.  We consider these proposals as first approximations which can be refined or totally changed in light of empirical evidence.

However, one can criticize the model from an opposite point of view: it may appear too complicated to be efficiently related to quantitative data.  We also partially accept this criticism,  Our attempts to relate the initial distribution of opinions and uncertainties with empirical data from quationsaires... were not entirely satisfactory.  We also had some difficulty in attributing concrete values to some parameters of the dynamics.  In this respect, the threshold model is easier to deal with.

nevertheless, the richness and the interprtability of the different dynamical behaviors of the model seem to us good arguments in its favor.  The interpretability indiacates that the complexity of the model remains humanly tractable, and that the model helps to bring a new way of understanding complex social phenomena.  Is that not what we ask from a model?

"

In the first paragraph they are depending on the assumption that if the fundamental dynamics are robust, than the specific parameter values can't significantly subvert their results. Essentially, the parameter values are noise. i don't know how to support or falsify that assumption.

The other tow paragraphs are itneresting, im going to hold onto them, ill remember this as 'the modeler's defense' I'm not so sure its a good one, but its good to know, and even good to be able to defend.