Question about Segovia-Martin & Tamariz (2018)

[seannyD]

Document: SegoviaMartinevolang12 Contributors: @jsegoviamartin @mtamariz

Thanks for adding your EvoLang paper! I have a two questions:

1) What does "convergence" mean here? It's quite vague. There's a variable in the database called "shared language" that's used in many simulations, is this close? (e.g. https://chield.excd.org/document.html?key=dallasta06languageGamesNetworks) Or maybe you want to be specific e.g. "convergence: lexicon"?

2) It would be easier to interpret the links with some short quotes from the paper pasted into the "Notes" field. For example, at the moment, it says that content bias influences population connectivity - I can't quite understand this.

[jsegoviamartin]

I did not have much time to take a look at this in Evolang, but I'm really enjoying your idea.

We compute convergence in the communicative system as a reduction in variation within the set of variants used by the agents. The amount of variation is quantified by the notion of entropy (Shannon, 1948). In other words, rate of convergence refers to the rate of information aggregation (Mueller-Frank, 2013) (or how fast the agents agree to use the same communicative convention). The model doesn't necessary apply to lexicon. It is a general model to explain the spread of cultural conventions. I think "shared language" would work better than "convergence:lexicon". I also considered the variable "conventionalisation".

As for the relationship between "content bias" and "population connectivity", we have found in our simulations that content bias amplifies the effect of the connectivity. The more content bias, the more differences between different levels of connectivity. I was not sure how to express this in the causal diagram.

Please, let me know if this clarifies a bit your questions. I would be happy to discuss more about it.

Regards, Jose Segovia

https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail Libre de virus. www.avast.com https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On 23 April 2018 at 15:04, Sean Roberts notifications@github.com wrote:

Document: SegoviaMartinevolang12 https://chield.excd.org/document.html?key=SegoviaMartinevolang12 Contributors: @jsegoviamartin https://github.com/jsegoviamartin @mtamariz https://github.com/mtamariz

Thanks for adding your EvoLang paper! I have a two questions:

1.

What does "convergence" mean here? It's quite vague. There's a variable in the database called "shared language" that's used in many simulations, is this close? (e.g. https://chield.excd.org/ document.html?key=dallasta06languageGamesNetworks https://chield.excd.org/document.html?key=dallasta06languageGamesNetworks) Or maybe you want to be specific e.g. "convergence: lexicon"? 2.

It would be easier to interpret the links with some short quotes from the paper pasted into the "Notes" field. For example, at the moment, it says that content bias influences population connectivity - I can't quite understand this.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/CHIELDOnline/CHIELD/issues/190, or mute the thread https://github.com/notifications/unsubscribe-auth/AU-57uD0mvt1m1AHqF4CtZcOX9APqhv8ks5trd9xgaJpZM4Tf9Xi .

-- José Segovia Martín https://jsegoviamartin.github.io/ @jsegoviamartin https://twitter.com/jsegoviamartin

[mtamariz]

Hi Sean,

I can't remember what / whether I submitted yesterday, but I'm sure I wouldn't put it like this. I will talk to Jose, who may have changed it / added it himself, and then get back to you.

This is how I'd do this study:

"convergence" should be "time to convergence". it is related to shared language but it has a temporal component. It could alternatively be called "time to converge onto a shared language"

"population connectivity dynamics" affects "time to convergence" (it could be positively or negatively, depending on the connectivity pattern) "content bias" affects "time to convergence" negatively [more content bias, faster convergence] "allocentric bias" affects "time to convergence" negatively "egocentric bias" affects "time to convergence" negatively "memory capacity" affects "time to convergence" positively [more memory capacity introduces more variation and so slows down convergence, so more time to convergence]

The fact that content bias interacts with connectivity is beside the point in these causal maps, right?

All the best, Monica

On 23 April 2018 at 15:04, Sean Roberts notifications@github.com wrote:

Document: SegoviaMartinevolang12 https://chield.excd.org/document.html?key=SegoviaMartinevolang12 Contributors: @jsegoviamartin https://github.com/jsegoviamartin @mtamariz https://github.com/mtamariz

Thanks for adding your EvoLang paper! I have a two questions:

1.

What does "convergence" mean here? It's quite vague. There's a variable in the database called "shared language" that's used in many simulations, is this close? (e.g. https://chield.excd.org/ document.html?key=dallasta06languageGamesNetworks https://chield.excd.org/document.html?key=dallasta06languageGamesNetworks) Or maybe you want to be specific e.g. "convergence: lexicon"? 2.

It would be easier to interpret the links with some short quotes from the paper pasted into the "Notes" field. For example, at the moment, it says that content bias influences population connectivity - I can't quite understand this.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/CHIELDOnline/CHIELD/issues/190, or mute the thread https://github.com/notifications/unsubscribe-auth/ANkAjoZlr5JHif3z5mXDvfw-1waHXzMAks5trd9xgaJpZM4Tf9Xi .

[seannyD]

Thanks for the discussion. The amplification effect is a kind of interaction effect. There doesn't appear to be a standard way of representing interactions in causal graphs. Some have an edge which links a node to another edge (i.e. the node interferes with the effect of another link). That's not possible in CHIELD at the moment. The other option is to have a node whose label is the interaction between two variables (e.g. "content bias * population connectivity). I'm currently reading up about this, but see e.g. this paper's section on sufficient causation.

However, in many cases the issue can be resolved by thinking about the extra steps in between. For example, you could add an extra node called "rate of dispersal", then have:

content bias > rate of dispersal population connectivity > rate of dispersal rate of dispersal > time to convergence

"rate of dispersal" might not be an actual parameter in your model, but I'm guessing this is part of the explanation of why these things are connected. In the nodes you could add information that there's a complex relationship between content bias, population connectivity and rate of dispersal.

I'll leave it to you to resolve the coding of your study. If you go to the page for your document, you can click "edit data" and make all the changes you want.

Thanks again for being pioneers!