Mike Arauz is a strategist at The Problem with Duncan Watts' Research on the Theory of Influentials
Duncan Watt's research (download the full PDF here) has been getting a lot of attention around the blogosphere lately (Noah Brier, Gavin Heaton, and Valeria Maltoni) and I think that it warrants a big caveat.
The Theory of Influentials, as generally described in Malcolm Gladwell's The Tipping Point, and The Influentials by Jon Berry and Ed Keller, asserts that a small fraction of the total population of a given community have a disproportionate affect on the community's likelihood to adopt a new trend. As Watts acknowledges, this influence is due to an individual's reputation, knowledge on the given topic, or simply their many connections to other people in the community.
As described in this Fast Company article, Watts' findings seem to fly in the face of this assertion.
Watts discovered that so-called Influentials are no more likely than any random individual to start a viral trend.
In his paper, Watts writes that the variable in his model that refers to an individual's influence should be thought of as referring "not to how many other [people an individual] knows, but how many other [people an individual] influences with respect to the particular issue at hand." So it's not just how connected a person is, but a complex combination "of [an individual's] personal characteristics, subject matter expertise, authority with respect to the issue, and even the characteristics of the other individuals in [the individual's] community." However, an individual's influence is "mathematically equivalent to the notion of 'acquaintance volume,'" or how many people an individual knows.
So Watts is boiling all of that complexity that goes along with the theory of influentials down to a single number that has the same mathematical weight in his model as simply knowing a lot of people. Or, in other words, in Watts' research, an influential has no more sway over his or her peers than anyone else in the community; they simply affect more people.
So, if you believe that a small portion of a community is disproportionately persuasive in getting the rest of the community to adopt a new trend, regardless of that select group's reach, then Watts' research doesn't tell you anything.
I assert that each of us, as individuals, is connected to a complex social network – much more complex than Watts' model reflects. Some of our networks are larger than others, but our ability to be affected by other individuals in our community varies greatly from relationship to relationship. The variable ability of any other individual to influence our personal choices depends on our personal perception of that individual, how well do we know them? do we trust them? do we respect their opinion? And for each of us, there are a few individuals who carry a disproportionate sway over our inclination to change our position on a given issue. I'd like to see another study in which another variable – independent of Watts' influence variable – that reflects this aspect of social dynamics is added to the model.
In the mean time, we should all take Watts' study with a big grain of salt, before we start tossing the concept of influentials into the trash bin.
If you think I've misunderstood Watts' research, or if there's further research that I've missed, please say so in the comments.
UPDATE: Ilya at Adverblog makes a good point: "My own beef with the theory of influentials is its implicit assertion, in my understanding of The Tipping Point, that if you have influence over a certain social network, this influence will have the same weight over the entire range of topics."
The Theory of Influentials, as generally described in Malcolm Gladwell's The Tipping Point, and The Influentials by Jon Berry and Ed Keller, asserts that a small fraction of the total population of a given community have a disproportionate affect on the community's likelihood to adopt a new trend. As Watts acknowledges, this influence is due to an individual's reputation, knowledge on the given topic, or simply their many connections to other people in the community.
As described in this Fast Company article, Watts' findings seem to fly in the face of this assertion.
[Watts] programmed a group of 10,000 people, all governed by a few simple interpersonal rules. Each was able to communicate with anyone nearby. With every contact, each had a small probability of "infecting" another. And each person also paid attention to what was happening around him: If lots of other people were adopting a trend, he would be more likely to join, and vice versa. The "people" in the virtual society had varying amounts of sociability – some were more connected than others. Watts designated the top 10% most-connected as Influentials; they could affect four times as many people as the average Joe.
Watts discovered that so-called Influentials are no more likely than any random individual to start a viral trend.
In his paper, Watts writes that the variable in his model that refers to an individual's influence should be thought of as referring "not to how many other [people an individual] knows, but how many other [people an individual] influences with respect to the particular issue at hand." So it's not just how connected a person is, but a complex combination "of [an individual's] personal characteristics, subject matter expertise, authority with respect to the issue, and even the characteristics of the other individuals in [the individual's] community." However, an individual's influence is "mathematically equivalent to the notion of 'acquaintance volume,'" or how many people an individual knows.
So Watts is boiling all of that complexity that goes along with the theory of influentials down to a single number that has the same mathematical weight in his model as simply knowing a lot of people. Or, in other words, in Watts' research, an influential has no more sway over his or her peers than anyone else in the community; they simply affect more people.
So, if you believe that a small portion of a community is disproportionately persuasive in getting the rest of the community to adopt a new trend, regardless of that select group's reach, then Watts' research doesn't tell you anything.
I assert that each of us, as individuals, is connected to a complex social network – much more complex than Watts' model reflects. Some of our networks are larger than others, but our ability to be affected by other individuals in our community varies greatly from relationship to relationship. The variable ability of any other individual to influence our personal choices depends on our personal perception of that individual, how well do we know them? do we trust them? do we respect their opinion? And for each of us, there are a few individuals who carry a disproportionate sway over our inclination to change our position on a given issue. I'd like to see another study in which another variable – independent of Watts' influence variable – that reflects this aspect of social dynamics is added to the model.
In the mean time, we should all take Watts' study with a big grain of salt, before we start tossing the concept of influentials into the trash bin.
If you think I've misunderstood Watts' research, or if there's further research that I've missed, please say so in the comments.
UPDATE: Ilya at Adverblog makes a good point: "My own beef with the theory of influentials is its implicit assertion, in my understanding of The Tipping Point, that if you have influence over a certain social network, this influence will have the same weight over the entire range of topics."
Sunday, January 27, 2008
|
7 Comments:
Mike,
The overlooked piece in the article on Watts that I grabbed onto for my post is the idea of something going viral when the context it right for it to go that way. In other words, the context and not the person (influential or not) determined what will go viral. In that case, even someone with less visibility might have the *critical connections* to make it go viral.
I talk about application in viral marketing, not about tossing in the can the theory of influentials altogether. My other observation around influentials is that they will tend to focus on their area of comfort -- as we all do -- or expertise... so you're better off being everywhere your prospects are as well. Just a thought.
Thanks for the response, Valeria.
I think that there's a lot of truth to the notion that context is crucial.
I think the key is to reach the resonant core - the small group of individuals who will be most easily persuaded to adopt your idea. This would be similar to the network that Watts' identifies as optimal.
And then use the attention from that community to build momentum in increasingly larger and looser networks.
But, every individual is still influenced by their peers. And that influence is much more complex and nuanced than is reflected by Watts' Threshold Rule.
Bottom line is that this whole thing is very complex, and I look forward to digging deeper.
Also, Noah Brier has a thorough and thoughtful new entry up about this.
I did like thepart of Watts theory that emphasized chaos, and the experiments were very interesting, but I view it as an interesting compliment to Gladwell's approach vs. a substitute.
Good take here.
Unless I’m way off the mark here, and correct me if I am, the only debate here is whether you should spend your marketing dollars targeting your ads at a lower number of influentials or reaching a broader market. This is a debate about cost trade-offs, not the fundamental nature of social networks.
Given that the objective of most marketers is to spread a given idea in the most cost-efficient manner (and it is), given that improvements in technology will make it more cost-efficient to identify and target influentials (and it will), and given that influentials themselves will become more connected via social media tools (and they will), word-of-mouth/ social/ viral marketing practitioners will do well to continue to focus on the tipping point potential of influentials.
I agree that trends do not start with influentials. Trends start with the average Joe or Jane who watches a video and sends the link to his friends via AIM, reads a article and emails it to a friend, sees something on the street, snaps a photo of it and sends it to his/her friends via multi-media text message. It is the high school kid who comes home and spends 4 hours on AIM, MySpace, Facbeook etc that lights the spark to eventual fires. This happens across the board on so many levels.
At some point mid-way in this process it reaches the influentials who also see the uniqueness and coolness of the content they are viewing. They have the connections to spread it quicker, but it can definately done without them.
Mike,
Great post. As you've noted, there is a level of complexity involved here that may escape data modeling. Bottom line: not all influencers are influential on every topic, and reach does not necessarily equate with results. It's about connectedness, relationships, the power to persuade, and ultimately, the value of the idea or product that is being hyped.
Mike, I think you've missed part of the details of the math. Watt's model doesn't necessarily ignore that idea that a single node (person) may be more influential over a specific other node (person) than another, but it simply assumes that across any given person's entire network, those ups-and-downs of influence balance out. He could easily add a random perturbation effect to the model and it would show an identical result.
The question is whether you believe there are some people whose opinion(s) are more trusted across their ENTIRE NETWORK (on average) than other people's opinions are across their own networks. Basically, whether I have more sway over my "people" than you do over yours. If you believe this is true, then Watts' model doesn't make sense, as it appears that every node has the same chance to infect any other node as anyone else. But, if you don't really believe this is true (or that there are so few people that fall into this category as to be generally irrelevant), then Watts' characterization of an "influential" as someone that just has more connections makes perfect sense.
Personally, I think that there are some people who are just more trusted across their networks than others. However, I don't really know that these differences are significant enough to affect the model. Additionally, whether you think that Watts' model adequately accounts for YOUR version of what an influential is, it actually does what it is supposed to do -- adequately model what previous theories have spouted off as being an "influential".
Post a Comment
<< Home