# Different research fields are different and those differences change over time *This feels like it should be a Dr. Seussian Poem* Hot field cold field Big field small field Fast field slow field Some fields have many questions, some have a few. Some fields are old, some are new. * In some fields most papers have a single author, in others there are hundreds. * In some fields conferences are higher prestige than journals, in some vice versa. In some all the prestige goes to books. * In some fields, papers are three pages long, in others they’re 70 pages. * Papers capture everything you need to replicate in some fields and in others you need to talk to another lab’s technician for days to discover that he used nose-bridge oil to get the experiment just right. ([[Tacit Knowledge, Trust, and the Q of Sapphire]]). * Citations mean very different things in different fields — in some, it means “I am basically stealing this person’s idea” in others, it’s a signal that you’ve read the literature, and in others it’s a pointer to the originator of an experimental technique you use. Looking at how these differences change over time, it’s useful to draw on biological analogies ([[The knowledge frontier is a high-dimensional garden]]). Fields can grow, hibernate, shrink, bud, splice, and die. They need different care in different stages in their lifecycles. While anybody who has actually worked on the knowledge frontier implicitly knows this, the explicit way we talk about fields doesn’t capture it. That may have been ok when the implicit dominated, but[[The systemization of science since the 1950s treats all fields the same]]. [[Kanjun Qiu]] describes the evolution of a field as S-curve-like: “pre-field -> field -> incremental growth.” I would argue that each of these are states that a field can be in but that instead of there being a single natural progression, there is a ‘state machine’ that fields travel through similar to how [[The linear innovation pipeline is more harmful than useful]]. A simple example is a field like AI that has arguably gone through several cycles of incremental growth -> field -> incremental growth -> field. Obviously what constitutes a “field” is incredibly [[Nebulous]], but I don’t think you could find an atomic unit that always follows a single path. To argue against this point, if we treat fields as duals to [[Thomas Kuhn]]Ian paradigms, each resurgence of AI could arguably be a new field because while the big questions stayed the same, the rules around answering them changed. [[Paradigms outline a set of questions and the rules around answering them]]. ### Related * [[Fast moving fields have lots of unconnected nodes]] * [[Slow moving fields are almost “fully connected”]] * [[New fields need to be exclusive]] * [[Formalizing a field too early can lead to premature optimization]] * (ref)[[Why do people feel like their academic fields are at a dead end?]] * (ref)[[Jan Van den Akker’s unit of progress in a field looks like heuristic statements of a specific form]] [Web URL for this note](http://notes.benjaminreinhardt.com/Different+research+fields+are+different+and+those+differences+change+over+time) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Different+research+fields+are+different+and+those+differences+change+over+time)