Awesome

Carnegie Mellon University 10717: The Art of the Paper

Logistics

Instructor: Zachary Lipton (Director of ACMI lab)
Time: Tuesdays & Thursdays 4:40pm–6:00pm
Location: GHC 6115 (in person)
TA: Charvi Rastogi
Reading List: TBD

Course Description

Scientific claims, evidence, and intuitions are all communicated through writing. Positions on what science is worth conducting, and what findings worthy of publishing are all expressed primarily through writing. The product of a successful scientific career tends to be a collection of influential papers. Ultimately, superior writing skills are often what differentiates influential scientists and principal investigators from their peers. Strong writers are better able to convince the wider field of what problems are worth solving (and funding) in the first place, and to develop the didactic resources that propagate scientific knowledge across communities and generations. And yet, the craft of scientific writing remains surprisingly neglected in engineering fields at every stage of training. Writing is often lumped under “presentation skills'', and writing assistance is often remedial in nature. Yes, the role of papers is sometimes to present material expressed more directly elsewhere (e.g., in proofs or code). And yes, many budding scientists do require remedial assistance. But the role of scientific papers is far greater: they are the sole vehicle for expressing whole categories of scientific arguments, and failures of writing nearly always bely failures of critical thinking.

This class, offered in mini 4 (second half of spring semester) will tackle scientific writing from several angles. First, we will tackle general principles, including (i) the expression of both normative and descriptive claims; (ii) heuristics for composing scientific prose that goes down smooth; and (iii) nuts and bolts technical matters, including peculiarities of scientific grammar, the scientific voice (e.g., what we mean by “we”), and the {how, when, and why} of citations. Next, we will then tackle the genre of the ML conference paper, including (a) the section-by-section architecture of the archetypal paper (how to write a strong abstract, introduction, etc.); (b) strategies for communicating complex technical ideas to distracted (and often short-fused) reviewers; (c) the appropriate ways to articulate contributions and set expectations; and (d) heuristics for successfully navigating the rebuttal process. Finally, we will discuss principles that govern writing for more general audiences. The class will employ a mix of lecture material, large-group discussions, and small-group exercises. Students will be expected to bring writing samples from their own work (publication status doesn’t matter) and to be forthcoming but respectful when criticizing the work of their peers. The aspiration of this class is to help each student to improve not only as a writer, but also a reader and as a scientist.

Tentative Schedule of Classes

Deliverables