“Good teaching’s like jazz”
This statement from Solomon Russell’s special invited session has been rattling around in my brain since that talk. Given the context, I think he was trying to explain that good teaching means improvising around the needs of the students rather than following a prescribed lesson plan, but after thinking about the themes from the conference, I’d argue that this analogy applies as well if not better to data science.
To me, there are two main distinguishing characteristics of good jazz as distinct from most other genres: creativity and communication. Creativity refers to improvising within a structured set of chord changes (It’s not just me: according to legend, Jo Jones threw a cymbal at a teenage Charlie Parker when Parker’s solo on “I Got Rhythm” got completely divorced from the chords to the bridge). Communication refers to listening attentively to what the rest of the band is doing and improvising to match or “call back” to it. I’d argue that one can be an amazing musician without a great sense of creativity or communication - but one can’t be an amazing jazz musician.
And yet - aren’t feature engineering and model building fundamentally creative endeavors? Didn’t we hear at the conference that listening attentively to understand the problem that is actually to be solved is a critical data science skill?
If we include some of the themes from Mine’s keynote and the invited talks sessions, I can come up with a set of “five C’s” that I think make a pretty good synopsis of learning objectives for any data science program.
- Curiosity - the ability to ask interesting questions through which the data and its generating process are “skillfully interrogated”
- Complexity - the technical know-how to deal with the scale and messiness of the data and its generating process
- Creativity - the ability to riff on fundamental building blocks to “improvise” something appropriate to the problem (linear regression is the data science equivalent of the C major scale?)
- Communication - not just speaking and writing, but also listening and reading attentively to understand the problem actually to be solved, and listening skillfully to the data (in the sense of Solomon’s talk, “An Analysis by The Analysis”?)
- Community - in the sense of the invited talks session I chaired, using data science to inform and empower one’s local community and/or, in the sense of Mine’s keynote, our responsibility as data scientists to the broader human community.
And yet - if you summarized your program’s learning objectives with the five C’s of “Curiosity, Complexity, Creativity, Communication, and Community” - wouldn’t that describe any worthwhile program in the arts and sciences in the age of AI?
What do you think? Would you use or adapt these learning objectives for your program? What “C word(s)” am I missing?