TEDx Tuttle Presentation

Editor's Notes

• #2: Social media needs to be measured in different ways “Civilization and its discontents” where Freud talks about the conflict between the individual’s desire to maintain their individuality and society’s pressure to conform. I work as a marketeer; for years I’ve been part of that pressure. Social Media presents -- express their individuality -- new set of unconscious pressures.
• #3: People are talking about your products, messages, about your ideas, about your policies.
• #4: Think about how many brands, NGOs and political organizations are trying to get you to do this. ethically grey.
• #5: In fact, I think it’s naive.
• #6: Then we think about it as something we can brand. Or buy. I’d say that social media is something that’s emerging as a RESULT of the ease of publishing. TOO MUCH information.
• #7: Twitter Where do you hear about stuff FIRST? Unconscious subtle instantaneous decisions respect like love trust
• #8: Filters are almost fractally complex and individual. That’s why they WORK. So here are some of the ways we’ve tried to get around it.
• #9: how do I talk to as few people as possible to reach as many of the right people as possible? And so it’s natural that we’ve tried to take this model and apply it to the social media
• #10: people expose their social graphs on sites like Twitter and Facebook.
• #11: Here’s mine.
• #12: I’ve got 213 friends (actually I’ve been pruning my friends)
• #14: But this is quite a narrow way of looking at things. You don’t know anything about my friends. All you know is something like this...
• #15: Mat has some friends. This graph is a very simple EGONET
• #16: Here’s how many friends my friends had back at the tail end of 2007 On the popularity model like this you’d probably start off by talking to the people with most friends.
• #17: Here’s how many friends my friends had back at the tail end of 2007 On the popularity model like this you’d probably start off by talking to the people with most friends.
• #18: So there are better models -- like the Eigenfactor model -- where you look not just at HOW MANY, but WHO. Roughly speaking, the more POPULAR people that link to you, the more influence you have All sorts of things inc Google’s PageRank works along these lines, so this is sort of “Page Rank for People”
• #19: So in the case of my profile, you’d rather take on someone whose friend profile was more like THIS. Someone who knows more famous people as it were. But again, this is a bit simplistic.
• #20: Let’s talk about Betweenness. Sometimes we talk about people being “very connected.” They tend to have high betweenness.
• #21: Here are four people’s EGONETS back from the time when I began exploring. You can see who they know in common. Here’s me. If you wanted to get an introduction to all four people, you could save time by just talking to me. I’ve got what’s called high BETWEENNESS in this network.
• #22: People with high betweenness are important. Without them, networks fall apart and messages don’t get through.
• #23: In this network, the bright spots are the people with high betweenness. If you remove them
• #24: The network falls apart.
• #25: Here’s another example. This time its Republicans and Democrats on Twitter. If you take out only two betweeners,
• #26: You get a world where the Republicans don’t know what the Dems are saying and vice versa. Betweeners are people who span two or more worlds. Let’s take a closer look at why there aren’t MORE of them. It’s because of a thing called Homophily.
• #27: homophily means (more or less) that people tend to hang around with and trust people who are like them And that people become more and more like the people with whom they hang around. It’s a nice fat feedback loop.
• #28: Basically birds of a feather flock together. That also means that you can tell a man by the company he keeps. Which is interesting to us marketers. Humans are very GOOD at this sort of filtering. Or at least very PRONE to it. But how can we use it? Can we automate it?
• #29: Well -- it turns out we can. Up to a point. Here’s a map of Twittering MPs. The colours tell you what their party affiliation is. Only, I didn’t colour them by hand. I asked the computer to work out what the most likely cliques and factions were. It got one person wrong.
• #30: Derek Wyatt is actually Labour. But it’s not bad for a small data set.
• #31: And here’s the same exercise performed on the US Congress. We got six out of 46 wrong. Still feels pretty good. We’re beginning to be able to tell something about someone based on nothing more than WHO THEY KNOW.
• #32: Because what we MIGHT be able to do is to find people who are more susceptible to certain kinds of message. People who are already experiencing real social pressure to make certain decisions. We might be able to REINFORCE that pressure, rather than try and INITIATE. Use their existing momentum. Sort of Kung Fu Marketing.
• #33: But the betweenness and homophily things also begin to explain why “viral marketing isn’t really viral.” A lot of people who don’t work in this area tend to have an oversimplified idea of how viral works.
• #56: 11:43 Media140
• #61: Most of the people who retweeted are already tightly linked -- recognized more than ONE name on the retweet schedule. There’s a SOCIAL act going on here; not simply sharing information, but using the act of sharing information to reinforce their relationships w/in the wider network
• #62: So here’s what I think. People are using their social networks to create filters that are impermeable to stuff we don’t want to hear about. And as marketers we’re looking for ways to get through those filters. But if we’re going to understand how to do this, we need to find ways to measure things like respect, liking, loving, and trust. s