Choose a precise scientific problem
Start with a phenomenon, unsolved technical limitation, measurable variable or reproducibility concern.
IDEA DEVELOPMENT / AI RESEARCH STUDIO
Build a stronger research question.
AI-assisted, literature-grounded research idea generation is planned. Until then, use this evidence-first process to develop feasible STEM research directions.
WHAT WE'RE BUILDING
From a field of interest to a testable hypothesis.
The future AI studio will combine verified literature, your available skills and equipment, feasible methods, known limitations and links to the original papers. It will never guarantee that a topic is novel.
✧AI · LITERATURE · METHODOLOGY
How to develop a good STEM research idea today
Follow these steps independently, with a supervisor, or using an AI assistant while verifying its citations.
Review foundational and recent work
Read highly relevant original papers. Keep a table of methods, datasets, results, weaknesses and citations.
Identify an evidence-based gap
Look for conflicting results, missing benchmarks, untested conditions, difficult assumptions or small experimental samples.
Write a testable question
Specify what you will measure or prove, the comparison, scope and expected uncertainty.
Design methods and baselines
Choose models, experimental controls, evaluation metrics, statistical tests and repeatability checks.
Check feasibility before committing
List required data, software, equipment, permissions, cost, timeline, safety and supervisor support.
Run a small pilot
Test one dataset, prototype or analytical case to discover problems before a major research commitment.
Document and revise
Record provenance, failures, evidence against the hypothesis and feedback from qualified reviewers.
A BETTER AI PROMPT
Start with evidence, not random topics.
Copy this template into your preferred AI assistant. Replace the bracketed fields and independently verify every source.
Act as a cautious STEM research mentor. I am a [academic level] student in [field], with skills in [skills], access to [equipment, data and software], and [available time]. My interest is [specific topic]. First ask for any essential missing constraints. Then help me identify 3–5 testable research questions informed by verifiable existing literature. For each, provide foundational papers and recent papers with real DOI links (mark any source you cannot verify), what earlier studies actually established, possible knowledge gaps (not guaranteed novelty), proposed hypotheses, methods and baselines, data and equipment needs, estimated complexity, risks, ethical requirements, and a low-cost pilot experiment. Challenge unrealistic ideas. Never invent citations or promise publication.