About a month ago I read an blog entry from @alanlepo where he spoke about the importance of evidence, see Recommendations Done Right Via IBM Connections. It struck a cord and I promised I would share my thoughts on the subject. So here they are :-)
Some years ago I made a slightly tongue in cheek prediction that “Content is Dead”. Well, to be more precise I suggested that “content will be replaced with a cloud of shared insights (the knowledge base) backed up with concrete evidence (content). As the breath of knowledge increases and the trust network grows there will be increasingly less need to go back to the evidence. Knowledge will be the cloud sitting on evidence integrated seamlessly into applications”
Ok, this may sound completely futuristic — I blame all the science fiction I watched as a child — however as weird as it may sound there is growing evidence that might support such a position. The volume, variety, and velocity of content is already overwhelming and its not going to get better any time soon. We already frequently rely on analytics to help us navigate the information and make sense of it all and this is again likely to become more prevalent with the increasing adoption of bigdata analytics.
Even today, often unbeknownst to ourselves, our perception of the world around us is being influenced by analytics algorithms, and nowadays more specifically social analytics. At the most simple level, a search engine chooses what content we should read through its ranking algorithms. How many of us ever scroll to the second or third page of search results? But thats only the beginning and is the least concerning since we still end up with a document that we can choose to read or not. It’s when the content starts to become secondary that we really need to think about the importance of evidence. For example:
- Influence algorithms assign values that could potentially have a significant impact on someone’s reputation (if we let that happen, which I sincerely hope we don’t). If you were to ask them to show the evidence backing up their opinion you would likely be given some mumbo jumbo about algorithms, reach, amplification, blah blah blah, … followed by an argument about patent protection. Sounds a bit like the incomprehensible financial instruments that got us into our current economic crisis :-(
- Recommender systems such as expertise location similarly apply analytics to large volumes of content in order to identify that person best suited to help you. Again, what evidence do they present to help you understand if their hypothesis is correct or not?
- Q&A systems are also increasingly using analytics to derive an answer from large volumes of content. In this case we are frequently not going to the original source(s) but just taking the answer at face value.
- The same with the new generation of decision support systems, driven by analytics of large volumes of data, which advise you on how best to treat a patient, handle a client problem, or improve brand sentiment.
Now, I am an analytics person so I am clearly NOT suggesting that we kill all analytics. I luv the stuff. BUT… what I am suggesting is that we need to ensure that there is a greater level of transparency around the algorithms and that the underlying evidence is made available in a digestible way.
Thankfully most of the respected analytics vendors are taking this provision of evidence as extremely serious and are actively integrating such traceability into their applications. My colleagues in IBM Research are building this into the social analytics machinery that underpins IBM Connections, see “Recommending Strangers in the Enterprise”. IBM Watson is another example of a signficant breakthrough in analytics which places a premium on the gathering, tracking, and presenting of evidence as part of its comprehensive analysis system.
So to wrap up this post I would call out to my analytics brethren and ask that as we increasingly integrate analytics into the fabric of society we stop treating the consumers of that analytics as dummies and proactively look to ways through which we can make our algorithms transparent and share the evidence underpinning our analysis results.