Social network from third interview
Image via Wikipedia

A recent BMJ paper by Steven Greenberg entitled, “How Citation Distortions Create Unfounded Authority,” touches on a problem identified by Andy Kessler in his book, “The End of Medicine” — medicine isn’t scientific enough yet.

Greenberg analyzed citation patterns among papers surrounding the belief that beta-amyloid is produced by and injures skeletal muscles in Alzheimer’s patients.

What he found is that social citation dominates scholarly citation, and that citation patterns confirmed hypotheses rather than disputing them. This effect also occurred among funding requests and grants.

The most authoritative papers in the set included four papers from the same lab, two of which used the same data but didn’t cite each other. A whopping 94% of citations flowed to these four papers. Six of the 10 most authoritative papers didn’t provide primary data (five were model papers, one a review).

Six papers that were isolated from the citation map provided data that disputed the beta-amyloid belief. Only 6% of citations went to these papers.

Citation diversion — the practice of misrepresenting a citation’s meaning — occurred in three papers early on. Over 10 years, these three papers led to 7,848 supportive citation pathways. Overall, this explosion of support occurred throughout the set of papers on the topic — a seven-fold increase in supportive citations compared to almost no growth in citations to papers critical of the hypothesis. Greenberg refers to this as a “lens effect,” a magnification effect that distorts the underlying reality.

Greenberg also documents citation invention, where authors in subsequent papers make claims about beta-amyloid, cite papers, but the citations don’t point to papers that support the author’s claims.

Finally, he found the same effects at play in approved funding proposals, with citation diversion and invention used to secure grants.

Greenberg identifies four major problems streaming from social citation practices:

  1. the preference for positive results in the literature leads to a bias for advancing known hypotheses
  2. the passive-aggressive nature of academic medicine means that instead of attacking or even acknowledging critical claims, critical papers are just not cited — they’re orphaned out of politeness
  3. these two behaviors lead to a social citation cascade that can flood the science with a cascade of under-scrutinized confirmatory results
  4. funding for new research adds funders’ desires to pursue viable hypotheses, further cascading convincing theoretical narratives with apparent (albeit social) citation strength

For medicine to become scientific and not about stories, the kind of social citation captured here needs to be limited sharply. Evidence-based medicine isn’t the answer, since it’s often susceptible to the same pernicious effects, especially since it can only evaluate what’s been funded. Veering from narrative to narrative, and having these frameworks persist for years and years only to be wrong or ineffective, erodes trust in medicine and blocks advances toward scientific solutions.

Medicine is at its best when the science is simple and clear — the powers of vaccination, nutrition, infection control, antibiotics, chemotherapy, and most surgeries are beyond dispute, and lives are saved or extended daily by these interventions. But while the search for crystallized answers is underway, a system that rewards misleading, positively biased information cascades needs to be restructured to allow the science to emerge only when its ready.

This means changing the publish-or-perish mindset, putting incentives in the right spots, qualifying research outcomes more equivocally, educating authors and researchers on the perils of sloppy citation (and the lensing effects it can have), and doing a better job of detecting when citation invention or diversion is occurring.

Social network theory has a role in this. Beyond the appeal of Facebook, Twitter, and other social media tools, the effect social citation has on the expression of research, in medicine and in other fields, is another reason why publishers need to understand and learn to deploy social network solutions — to enhance collaboration, and to eliminate misleading information cascades.

Social citation is the “socialized medicine” we should be worrying about in scholarly publishing. But can it be reformed?

Reblog this post [with Zemanta]