Richard Your abstract looks very interesting. I was wondering, for the kinds of network in your questions(financial markets, energy systems and social networks), how do you decide or find out what the connections and so structure of the network is? Alison
Thanks - yeah, good question. Short answer, this is theory, I havnt measured anything in the wild. Long answer, [thinking about it as I go] *theoretically* what matters is the positive feedback between behaviours (on nodes) and interactions between them (via connections) that modify those combinations of behaviours. There may be many different behaviours and many different types of connections and thus many network topologies - depending on what you want to pay attention to - so I appreciate your question. But the connections that matter are the ones that alter the coordination of the behaviors we are studying, and the behaviours that matter are the ones that alter the connections we are studying. They define each other. A bit abstract perhaps... lets try an example. So people have many different types of online interactions - e.g. online friends they use Zoom with and people they text message. These make different topologies, which topology is relevant? Depends what behaviour we are interested in... e.g. panic buying toilet roll. Talking to other people about this behaviour may influence how I shop and may influence my connections - who I text and who I zoom with. e.g. 'Im dropping out of this group, they're all idiots' or 'Im gonna stay plugged in here, these people have good intel', 'Im going to text Mary, she's sane'. This kind of change in topology could apply to either type of online relationship - so measuring it will be complicated. But the change that matters is 'what is the power of the influence between this behaviour in person A and this behaviour in person B before the change, and after the change?' - ie. the change in connection strength is defined by the change in the strength of this influence (change in mutual information, perhaps? Friston would know, probably). My theory says that if the connection is plastic it will change in the direction that makes the combination of behaviours they are doing now more likely to co-occur in future. e.g. Im panic buying, you are panic buying -> delta connection is positive, meaning the influence of us on one another is now more likely to make us do similar beh in future (either both panic buy again tomorrow or both dont). Why? How can this assertion be so general? Well, it could be be the opposite of course - but that would require me to seek stress - to change relationships in a way that makes me less at ease doing what Im already doing. People, like springs, tend not to work that way... put differently, you have to do some extra explaining (introduce other forces) to account for the alternative scenario. That's as far as I got.... there's always the short answer if you prefer :)
Richard
ReplyDeleteYour abstract looks very interesting. I was wondering, for the kinds of network in your questions(financial markets, energy systems and social networks), how do you decide or find out what the connections and so structure of the network is?
Alison
Thanks - yeah, good question. Short answer, this is theory, I havnt measured anything in the wild. Long answer, [thinking about it as I go] *theoretically* what matters is the positive feedback between behaviours (on nodes) and interactions between them (via connections) that modify those combinations of behaviours. There may be many different behaviours and many different types of connections and thus many network topologies - depending on what you want to pay attention to - so I appreciate your question. But the connections that matter are the ones that alter the coordination of the behaviors we are studying, and the behaviours that matter are the ones that alter the connections we are studying. They define each other. A bit abstract perhaps... lets try an example. So people have many different types of online interactions - e.g. online friends they use Zoom with and people they text message. These make different topologies, which topology is relevant? Depends what behaviour we are interested in... e.g. panic buying toilet roll. Talking to other people about this behaviour may influence how I shop and may influence my connections - who I text and who I zoom with. e.g. 'Im dropping out of this group, they're all idiots' or 'Im gonna stay plugged in here, these people have good intel', 'Im going to text Mary, she's sane'. This kind of change in topology could apply to either type of online relationship - so measuring it will be complicated. But the change that matters is 'what is the power of the influence between this behaviour in person A and this behaviour in person B before the change, and after the change?' - ie. the change in connection strength is defined by the change in the strength of this influence (change in mutual information, perhaps? Friston would know, probably). My theory says that if the connection is plastic it will change in the direction that makes the combination of behaviours they are doing now more likely to co-occur in future. e.g. Im panic buying, you are panic buying -> delta connection is positive, meaning the influence of us on one another is now more likely to make us do similar beh in future (either both panic buy again tomorrow or both dont). Why? How can this assertion be so general? Well, it could be be the opposite of course - but that would require me to seek stress - to change relationships in a way that makes me less at ease doing what Im already doing. People, like springs, tend not to work that way... put differently, you have to do some extra explaining (introduce other forces) to account for the alternative scenario. That's as far as I got.... there's always the short answer if you prefer :)
Delete