Capturing diffusion in action

In social marketing, diffusion is the hidden premise in nearly all our campaign strategies.  After all, we set population-wide objectives for campaigns only funded enough to reach a tiny fraction of society at large.  So if only 1-in-10 or 1-in-100 of our target audience actually sees our message — and I’m being optimistic here — how could that influence the other 9 or 99?  The implicit assumption is usually that our message will ‘diffuse’, i.e. spread onwards and outwards through word of mouth.

There’s a theory that supposedly describes how this happens: Rogers’ (1962) Diffusion of Innovations theory.  He studied how farmers came to adopt new technology for automated farming, and why some picked it up really early and others waited a really long time.  He graphed the adoption curve and divided it into segments: innovators, early adopters, early and late majority, laggards.

The implication for someone who wants to start a trend is that you need to identify the early adopters in a population — people who have LOTS of social connections — and convince them, and then the other segments will follow like dominoes.  This was certainly the message of Malcolm Gladwell in his book The Tipping Point, and both Gladwell’s and Rogers’ theories have turned up in massive RCT studies of community-level HIV prevention interventions, like Project Accept (Khumalo-Sakutukwa et al, 2008).

The problem is both theories are wrong.

Rogers just takes the normal curve and labels the standard deviations A, B, C, etc, and then assumes because A before B therefore A causes B.  Post hoc ergo propter hoc.  Gladwell assumes spread is determined by characteristics of well-connected, charismatic people (like himself).  But as Duncan J Watts demonstrated in real time using 16 separate bunches of real people downloading, sharing and rating digital music tracks, it’s not about particular well-connected individuals, it’s about having a large proportion of easily influenced people that matters (Watts, 2001).

To use a disease metaphor, we’re not looking for superspreaders like Typhoid Mary — it’s about what proportion of your population are vulnerable (ie. not immune) to the message you’re trying to spread.  This poses all sorts of really interesting questions for campaign planners.

One is:  what would information immunity look like?

And another: how would I track diffusion of my idea?

At a very superficial level, as Tyler Cowen Horan has shown, it’s possible to measure diffusion of links or hashtags on social networks like Twitter, using automated content analysis tools like OpenCalais and the publically available feed of all tweets for a given keyword.  That will give you a quantitative and social network view of the spread of a particular marker of an idea.  But it won’t give you any clear sense of how people are making sense of it.  That first question — information immunity — complicates the hell out of the second one.

So I was interested to see friends on Twitter discussing changes to the Healthy Kids Check as reported in a Fairfax news article:

Preschool mental health checks by Jill Stark (SMH, 10 June 2012)
THREE-YEAR-OLDS will be screened for early signs of mental illness in a new federal government program that will consider behaviour such as sleeping with the light on, temper tantrums or extreme shyness as signs of possible psychological problems.

In particular, one friend had a problem with the idea that GPs might be using a 3yo wanting to sleep with the light on to diagnose a mental health disorder.  A few different people engaged on three things I discuss here, and there was fruitful discussion about these aspects of the proposal, but my friend kept coming back to the 3yo wanting the light on.  This particular example clearly resonated with this friend’s strong concern about the pathologisation of natural processes in parenting and child development.

For my part, I noted that it’s about identifying known precursors or risk factors for future mental health conditions, rather than the GP diagnosing a current disorder.  This enables a ‘watchful waiting’ approach or possibly a referral to a child mental health expert for further investigation.

My own receptiveness (lowered information immunity) to the idea was increased by having recently read, shared and discussed on Facebook an article about ‘high reactive babies‘ — a temperament with a low threshold for alarm at small changes in environment, routine, noises, smells, that predicts anxiety in adulthood.  This resonated with me personally, as a colicky baby who grew into an anxious adult with a touchy tummy and insomnia.

This research isn’t mentioned in the article but it’s clearly a really important piece of background knowledge needed to interpret the new information; without it, it’s hard to understand why a 3yo wanting the lights on is relevant to anxiety the disorder, rather than just anxiety as a normal and natural emotion known to be experienced by 3yo children in the dark.  From a diagnostic point of view, it’s not the status of the light, on or off, that matters — it’s about the intensity of the anxiety and the strategies the child uses to manage the situation and their feelings.

Another key piece of background knowledge would be knowing that diagnoses are made based on the overall picture — the constellation of elements — whereas a screening tool developed with specialist referral as the intended endpoint will tend to pick out one or two key factors that are present in most cases of the disorder.

Both aspects are presented in the article, but they are laid out sequentially in quotes from different experts.  This is journalistically unexceptionable: one is always going to have to precede the other, and it makes more sense to present concern before response, problem before solution.  The problem lies in how it presents all three quotes as equivalent in authority:

    • Prof Allen Frances, chair of the DSM-IV task force, author of a book titled Am I Okay? (Expressing concern about overdiagnosis of ADD and autism resulting from ambiguous definitions in DSM-IV)
    • Anna Sexton from East Brunswick has children aged 3, 5 and 6 who all sleep with the hallway light on (Concerned this behaviour will be viewed as abnormal)
  • However, Chris Tanti, chief executive of headspace, the youth mental health foundation (Says early intervention did not automatically lead to children being labelled; Quotes figure that only 19% of clients showing signs of mental illness end up with a diagnosis)

I guess you could argue that Chris Tanti gets the last word, but unlike in oral argument (think of the closing statement in a trial), in text there’s no guarantee the reader even finishes the article.

And thinking in terms of information immunity, let’s compare which ‘accounts’ (ways of telling the story) will seem most familiar (a key measure of susceptibility to persuasion) to an audience of parents:  they’ve read lots and lots of articles about ADD and autism and overdiagnosis, and Prof Frances appears to be arguing against his own interests as a diagnostician(+10 credibility points!); Anna Sexton has read the same articles and shares the concern of the credible Professor, so she’s a caring and informed parent;  whereas Chris Tanti is using numbers, talking about a scenario most parents don’t want to imagine ever applying to their kids, and his argument seems to suit his professional interests.  In terms of who the parent-reader sympathises with, it’s two against one.

I’m not taking a swipe at Jill Stark.  This is a conscientious, intelligent, balanced and cogently presented article about a really important policy proposal.  Setting up that closing ‘conversation’ between the quotes about overdiagnosis practically guarantees it will be discussed by parents.  It does better at provoking diffusion than most campaigns in my own field, which often foreclose on discussion by finding and presenting a ‘single [simple] minded proposition’ that nobody can argue with (and so nobody does).

But I’m worried by how that discussion played out in my Twitter stream.  I’d like for Prof Frank Oberklaid, who mentioned sleeping with the lights on, to have some way of knowing that example backfired.   I’d like for the people developing the policy to know there’s concern about a key question — who does the diagnosis, if any, the GP or a specialist?  These are qualitative questions whose answers could feed back into better communication approaches and better health policy, if only there was some way to capture them…

A modest proposal

Imagine you’re about to send out a press release about a new campaign or policy initiative.  You know it’s likely to provoke discussion, perhaps involving issues that are tricky to deal with in a standard health news article format.  You’d like to see the discussion without engaging in surveillance of semi-private spaces like @-response discussion threads on Twitter.  And you’d like the ability to answer key questions arising in the discussion — or at least to signal that you’ve heard and acknowledged them being raised.

This is not a technically difficult problem to solve.  Let’s get a consortium of health communication practitioners and researchers and journalists together; Croakey would be perfect for this.  The consortium builds a website where people launching campaigns or policy initiatives can register and create a landing page with a short URL, which they offer to health journalists to include in their articles.  Readers can follow the URL to a page that lets them ask a question and receive a notification when an answer has been posted.  They can also see other questions that have been asked, and vote on them if they had the same concern.  Answers should be written by experts involved in the policy — they need to be substantive, as people can easily see through tightly-controlled PR responses.

I’d focus on Q&A and voting, as I’ve come to doubt the discursive utility of comment threads  in this hyperpartisan era.  And I’d want to let health journalists in on access to the usage metrics it would generate — after all, they need feedback too on what issues matter and how different article structures influence understanding and discussion of their work.  In this age of the “Australian Vaccination (sic) Network” and climate change “debate”, the need for post-publication opportunities to answer questions and correct misconceptions is screamingly clear.

Tender, loving financial incentives

This week, an American public health practitioner posted on the US national AIDS blog about a new program starting up in the States. The program is called TLC-Plus, short for “Enhanced Testing, Linking into Care, Plus Treatment” for PLHIV. Here’s how the author, Carl Dieffenbach PhD, explained the need for that middle component, “linking into care” —

“Unfortunately, many people don’t make it to their follow-up medical visits for a variety of reasons. Yet these individuals can continue to spread the virus in their communities until we entice them to step through the door, whether it’s giving them a pat on the back or offering financial incentives.”

Continue reading