Sunday, February 14, 2010

Pulling the information together

Search engines can be strange things. They are principally there to help users, with a gap in their knowledge, gain the information they need. This information may be a phone number, address, opening times, drug interactions.

Needless to say my interest is in clinical uncertainty.

Imagine if search engines were never invented and someone sat down to design a tool that would answer clinical questions. Would anyone really suggest that someone types in a few related terms, hits 'search' and then gets presented with 10-20 results that may have the information they're interested in? It's laughable really. This is further complicated by the fact that most of the clinical questions we've been involved in have needed more than one reference to answer. In our analysis of 350+ dermatology questions the average number of references used was 2.2.

So, the information is in disparate locations and hidden in a mass of other paragraphs (typically the information required is a paragraph or two located somewhere in a document with perhaps 100 paragraphs).

To my mind the ideal solution would be a user, typing in their full question (e.g. what are the causes of raised vitamin b12? as opposed to raised b12) and then they receive a brief response straightaway.

As far as I can tell this is a long way off (although I have seen some half-decent attempts recently). The Q&A services that we run (e.g. TRIP Answers) is another approach but it's relatively labour intensive. I quite like the approach that Aardvark is using which uses humans to answer the questions. As this wikipedia article states:

"Aardvark is a social search service that connects users live with friends or friends-of-friends who are able to answer their questions. Users submit questions via email or instant messenger and Aardvark identifies and facilitates a live chat or email conversation with one or more topic experts in the asker's extended social network. Users can also review question and answer history and other settings on the Aardvark website."

My business partner at TRIP (Chris) a GP says he already knows many of the answers to the questions we receive. He's always said that if he knows the answer he could return the response in 5 minutes (as opposed to the 60-120 minutes for a relatively easy Q&A for an information specialist).

Perhaps the Aardvark approach is the future - couple people with uncertainty with people who likely know the answer. It requires goodwill, but there's plenty of that about!

2 comments:

Carlos Cuello said...

Hi, Jon

I´ve been using Advaark for the last three days and their procedure is amazing, looks friendly and fast. Immediately I had the same idea you post here: why not doing it in medicine? The first problem might be the same we face with wikipedia: reliability of the answers. Would they be evidence based? How can be assure that the respondent is an expert in any field?

Cheers

Jon Brassey said...

Hi Carlos,

I think there are ways of demonstrating online 'reliability' (which isn't necessarily the same as 'evidence based). For instance you can allow other people (including the person that asked the question) to rate the response. If everyone says the response is great you're more likely to believe future responses. Conversely, if they all say the answer is rubbish - you might tend to disregard the response.

I really like the stack overflow model - see the section on reputation http://stackoverflow.com/faq