How To Choose a Good Scientific Problem

Three-page Forum essay by a systems biologist whose opening observation earns the piece its place: problem choice is essential to science yet “not usually discussed explicitly within our profession” — the vacuum gets filled by publication-count defaults. Alon’s frame is deliberately values-first: a lab is “a nurturing environment that aims to maximize the potential of students as scientists and as human beings,” and problem choice flows from that premise. What makes the essay durable is that the soft frame delivers hard tools — a ranking scheme, a timing rule, and a schema correction — each teachable.

The tools

• The feasibility × interest diagram (Fig. 1, viewed): problems ranked on two axes — feasibility (expected time to completion, a function of the lab’s skills and technology) and interest (increase in verifiable knowledge, “distance from the known shores”). Dominated problems get erased, leaving a Pareto front; where to sit on the front depends on career stage, and the figure labels the positions — first problem for a beginning student bottom-right (easy, fast feedback, confidence), postdoc top-right (time is limited), long-term lab plan at the grand-challenge end, decomposable into good smaller projects. A warning attached to the feasibility axis: problems easy on paper are hard in reality; problems hard on paper are nearly impossible.
• The three-month rule: no commitment to a problem before three months of reading, discussing, planning — “focused on being rather than doing” — with a celebration marking the transition to research. The justification is the asymmetry: projects take years, so a week spent choosing saves months. This is the operational answer to hasty first-idea commitment, and Alon is candid that it requires support and isn’t available to everyone (grant deadlines, funding).
• The inner-voice tests, for the subjective interest axis: “If I was the only person on earth, which of these problems would I work on?”; which ideas keep returning over months or years (versus recent-days ideas); how it feels to describe each project to an acquaintance. The claim: self-interest propagates — “the more you interest yourself, the larger the probability that you will interest your audience.” Self-expression is the deep version: each researcher’s interest pattern expresses a personal filter of values, and mentors can map it (visual aesthetics or abstract ideas? supporting dogma or undermining it?) to match projects to students.
• The nurturing schema and “the cloud” (Fig. 2, viewed): against the paper-shaped schema (A → straight arrow → B), the realistic schema meanders, loops, and passes through the cloud — a named phase where assumptions break down and emotions go negative — before often arriving at problem C, sensed in the materials at hand and better than B on both axes. Holding the wrong schema makes deviation feel like failure; holding this one makes the meandering part of the craft and licenses the switch to C.

Assessment

• Durable: the two-axis Pareto framing with life-stage weighting; the three-month rule as a scheduled selection procedure; the inner-voice tests; the cloud and problem C as vocabulary for the mid-project reframe. All of it teachable, which was the stated goal.
• Era-bound: the wet-lab/PI framing and the journal-culture backdrop; the mentor-student mechanics assume an academic lab structure.
• Caveats: evidence is one lab’s practice and the author’s experience, offered as such (“collected here and again offered as a gift”) — a values essay with tools, not a study; like hamming1986 it generalizes from survivors.
• In this library: the direct procedural complement to hamming1986 — Hamming argues selection and reframing are trained, scheduled activities; Alon supplies the schedule (three months), the ranking scheme (Pareto front), and the emotional instrumentation Hamming lacks. Alon’s problem C is Hamming’s problem transformation met mid-project — both locate the real problem as something discovered during the work, not before it. And the interest axis’s inner/outer voice distinction is a selection-side analogue of keshav2007‘s exit decisions: allocate scarce attention deliberately, against explicit criteria, at explicit checkpoints.