Chernoff faces are an old idea that have become more of a joke (eg https://dl.acm.org/doi/pdf/10.1145/3170427.3188398 ). The problem is that they are weird enough to occasionally lure in well-intentioned people who are new to datavis. Encoding abstract data as faces results in unpredictable, unrelated, gross visualizations that are difficult to read. The only good one I have seen is https://projects.propublica.org/graphics/workers-compensatio... and it’s creepy.