The web transformed the role of knowledge by making it instantly available but not inherently reliable. It became more valuable than ever to be able to locate it, assess it, and put it together with other pieces. But when we could see how deep the social practices that created knowledge run, we wanted more transparency from all who are making knowledge claims. We’re still figuring out how much to make transparent, how much we want to accept as settled conclusions, and how we want to invite our customers and the general public in on the conversations. If you called this a phase of knowledge “un-management,” I wouldn't complain.
And now machine learning is advancing knowledge by undoing its most basic underpinnings. For thousands of years, the point of knowledge was to provide reliable building blocks for new ideas and new knowledge. But machine learning works by scooping up tons of data, understanding that there are likely errors and outliers in it. Useful machine learning models are forged by statistical engines that do a good enough job finding what’ s relevant. However, when data reflects historical biases, machine learning's algorithms can easily elevate those biases, while hiding how that happened in inexplicable complexity. Both of these outcomes are due to a fundamental shift in knowledge—from what we can rely on to data that can tell us useful things that are reliable enough for our purposes.
This is good news for KM. If knowledge stayed the same, where would the opportunities be? Or the fun?