Research in “the cloud” and the honeymoon phase

A few days ago, I watched a TED talk by physicist Uri Alon on the emotional experience of scientific research. It is a fun talk, and I thought the message is important.

See the video here.

Uri argues that science is presented in a manner that poorly sets up expectations for scientists: the presented story is that we scientists brilliantly devise a hypothesis, run some experiments, and verify this hypothesis. Then we hit print and science is done. Instead, Uri sheds light on the actual process, which involves a taxing trek through what he calls “the cloud.” The real story is that a scientist has a hypothesis, tests it, finds that it is wrong, and then enters this cloud, which is a place of uncertainty, despair, and confusion. The scientist has to challenge his or her basic assumptions, and only by doing so, will be able to produce a new hypothesis that truly advances human knowledge.


Though Uri is a physicist, I think this message is just as relevant for computer science. I can’t speak for all fields within computer science, but at least in machine learning research, this is almost always the way progress occurs.

One problem of applying Uri’s message to computer science is the nomenclature. The cloud (or is it the Cloud?) is a buzzword that is already somewhat defined (but not very well). So using it in this context could lead to confusion (which is a bit too meta to be useful). We need another term for the computer science version of the cloud. Or we could just use it and stop using “the cloud” to describe the Internet, because we already have a word for that.

I would add to the process Uri describes another important concept to prepare researchers for the emotional roller coaster of science: the honeymoon phase.

The honeymoon phase is what I call the period when I’ve come up with an idea, perhaps it’s an algorithm, a theorem, or an application, that I think will work. As I start incrementally testing the idea–coding up the algorithm, for example–it starts to seem more and more correct. A subroutine of the algorithm works exactly as it should on paper! The learning algorithm correctly learns from a synthetic data set! If we assume a lemma, the proof actually works out! These small victories invoke a sense of euphoria and often come with daydreams of how easily this new research will lead to publication and recognition.

In reality, the honeymoon phase is almost always followed by a discovery that something is wrong, which leads to a sharp turn directly into the cloud. This contrast from the highs of the honeymoon phase to the lows of the cloud is jarring.

Like the message from the TED talk, I believe acknowledging that this sharp transition is part of a shared experience could help limit the emotional distress cause by the process of research. I’m not sure if there is any particular strategy for intelligently handling the highs of the honeymoon phase better, and I’m hesitant to suggest to anyone not to enjoy it while it’s happening.

Next time on Terrible Emotions in Science: Rejection…


One comment

  1. Pingback: On Rejection and Acceptance Rates « Bert Huang's Blog

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s