The discussion centers on the importance of a bottom-up approach in systems design for data science, as advocated by a participant with extensive experience in data handling. The participant emphasizes the challenges of maintaining a fresh perspective on data problems due to their expertise, which can lead to assumptions about others' knowledge. They also touch on the potential pitfalls of a top-down design approach, particularly in hypothesis testing, and the need for balanced documentation to avoid user frustration.
PREREQUISITESData scientists, systems designers, and anyone involved in data-driven decision-making who seeks to improve their understanding of system design approaches and documentation practices.
Jameson said:Interesting article!
I work with large amounts of data every day and would consider myself an expert in the field when considering the whole continuum of "Can't use a computer" to "Coding at Google". I grab data, clean it, join it, model it, and present it.
That is important to your point, because what I lose in my years of training is the ability to see data and a problem with new eyes. Once you know something, you can try to empathize with the mind of someone who doesn't know it but it is quite difficult.
Jameson said:Even worse, you can be very unaware of the facts that you internally assume everyone must know yet they don't.
So I fully agree with you that systems benefit from building bottom-up, not top-down.
Jameson said:Some guidance is helpful but too much will leave the end user clueless and frustrated.