So, you want to go out for dinner. But where? If you are in London, UK, there are more than 20,000 restaurants. Cue Google ‘best restaurant’? I just tried it and, thanks to IP targeting, I got everything (local) from ‘the pub’ to a multi-course tasting place with wild garlic broth and broccoli. To find my actual ‘best restaurant’, Google would need to take into account whom I was dining with, my budget, preferences, allergies, tendency to dance on the table after a good night out…
In the world of identifying Key Opinion Leaders (KOLs) for pharma, there are accepted methodologies but no ‘fixed’ assessment criteria, and indeed endless ‘choices’ to select from. So, how do you find the optimal partners for your business?
When I started in KOL mapping more than 20 years ago, there wasn’t a huge amount of data available to help make your choices. Often a medical writer would simply review publications and pick a selection of authors for the ‘KOL list’. As little data were available from many regions of the world, their KOLs were largely unrepresented.
Times have moved on, and with them the explosion of data in every facet of our lives; both personal and professional. According to Forbes, 90% of the world’s data was generated over the last 2 years.
This new richness of data has enabled us to move from a serendipitous to a more objective way of discovering an entire range of KOL types (from specialists to primary care to regulatory to digital influencers), their activities and opinions through their online profiles and activities. However, you can quickly be in a situation of information overload. Developments in analytics and big data, as well as AI, Machine Learning and NLP technology have changed KOL mapping methodologies completely and helped us to wrestle with the volume of data and speed of processing. But how do you ensure you are evaluating the right data in the right way to produce meaningful results?
As well as AI’s and other technology’s reliance on historical information, there is a significant risk of bias and misinformation. Humans are inherently biased, and so are our creations; AI included (see this interesting recent article in Nature highlighting AI’s diversity crisis). The difference is that intelligent humans can identify and compensate for our biases, whereas AI sees whatever data is programmed into it as objective truth. This is why social media companies use human moderators and fact-checkers to limit the spread of misinformation. Without this human stop check, the whole system crashes.
At Scout, we see ourselves as the human data company. We understand that the role of data has changed beyond recognition, but we also recognise that in some cases this has been to the detriment or loss of human involvement. We feel the latter is as important as ever and will continue to be so.
Thinking about how KOL mapping has evolved, even a few years ago, a ‘global’ KOL list based on analysis of available data would have been dominated by US and European male KOLs over the age of 50. We would hope this is no longer the case. However, if we think about how AI works, we recognise that it analyses and presents a helpful view of existing data. AI would see the people who have historically had a high profile and influence. Those who have historically been selected to present at conferences or have had their papers published, for example.
In a recent JAMA analysis of nearly 9,000 panels and invited lectures, with 23,000 speakers across 98 conferences around the world, it was found that only 30% of speakers were women and 40% of panels were all-male. Women are still suffering from historic gender bias impacting on their exposure, and it takes people who notice and care and speak up, to shift the narrative. This is of course also sadly true with regard to the inclusion of groups from different ethnic and cultural backgrounds.
At Scout, we work with our clients in an actively inclusive way to uncover and leverage influencers from a full range of backgrounds with myriad different skills: this will include individuals with a lot to offer but who may not have as much data to their name. Our success is due to truly understanding the data sets and their validity – what data can be found and equally important what we know will be missing. Our KOL lists are diverse in all senses: skills and expertise, particular interests within a therapy area, gender, age profile, cultural background and geography.
Algorithms and AI are helping us in so many ways; to harness the power of infinite data sets to drive forwards our understanding and efficiency. However, there is a danger in over-reliance and underrepresentation of human assessment that clearly reaches far beyond the world of finding KOLs to partner with pharma. 5% human 95% AI isn’t right to me. It needs to be a balance. It’s people that move things forward, not AI. At Scout we celebrate that.
By Lucy Walker, Operations Director, Scout
Lucy.walker@scoutdata.co.uk
