Over the years Gmail’s autosuggest has learned how I write with terrifying accuracy, creating a situation in which significant chunk of my routine e-mails are algorithmically generated. I suspect this exaggerates my written quirks, ticks & deficiencies by feeding them back to me. Moving to Outlook has filtered my writing through an auto-suggest which hasn’t learned my style. The result is that I’m writing more effectively, particularly with regards to brevity and grammatical expression.

There’s a lesson in this about everyday dependence on ML systems. It’s left me aware of how my written communication has been filtered through (a) machine learning which feeds my own stylistic quirks back to me (b) primarily communicating within closed peer-to-peer collaboration networks. This has been a useful push to write more carefully as I begin to interact with a wide range of new colleagues, as well as a large cohort of students. In this sense ML can be a spur to reflexivity about our communication, as well as something which gets in the way of it.