AI Research Partner

AI takes part in every step from phase 1. Every AI contribution is labeled, and researchers make the final call.

AI demo · sample response The AI assistants are currently a demo: sample responses built by inserting calculated results into fixed sentences. Live LLM integration comes after phase 2. Every demo response carries a dashed 'AI demo' badge.

Researchers experiment and decide. AI checks and analyzes.

01 · ResearcherRun the experimentExperiment and measure with the common protocol.
02 · ✦ AIData quality checkFinds unit errors, calibration range issues and outliers.
03 · ResearcherReview and shareFix what was flagged and publish the data.
04 · ✦ AIPatterns and interpretationExplains results and compares similar experiments.
05 · Researcher + ✦ AI suggestionDesign the next runAI suggests conditions; the researcher decides.

Researcher step✦ AI stepAs data from around the world grows, AI adds predictions and experiment suggestions.

The AI research partner working with you now

Experiment design assistant

Suggests protocol-compliant conditions and missing controls.

Researcher: Decides whether to adopt

Phase 1 · demo

Data quality AI

Flags unit errors, calibration range issues and pH effects.

Researcher: Fixes values or records a reason

Phase 1 · demo

Interpretation assistant

Explains graphs and reaction rates in plain words.

Researcher: Reviews before reporting

Phase 1 · demo

Literature assistant

Tabulates conditions and results from papers with source locations.

Researcher: Checks against the original

Phase 2+

Multilingual assistant

Answers protocol questions and translates discussions.

Researcher: Checks the original text

Phase 2+

Start analyzing

AI capability stages

Stage	What	Available	Label
Calculation (not AI)	Concentration, removal, k and R², control comparison	Phase 1	Calculated
AI research partner	5 LLM-based assistants	Phase 1 (demo now)	AI suggestion · researcher check
Data assist AI	CSV column mapping, outlier detection	Phase 2	AI assist
Prediction model	Estimate k and uncertainty from conditions	After release criteria	Prediction · experimental
Experiment suggestions	Suggest conditions worth testing next	After external validation	Candidate experiments

AI collaboration principles

Verified code calculates numbers like removal and k. AI never makes up numbers.
Every AI output carries a ✦ AI badge; researchers adopt, edit or reject it.
AI answers show their evidence (dataset ID, protocol section, paper).
Cases where AI was wrong are recorded so everyone can see its limits.

Prediction model release criteria

Prediction stays off until every criterion below is met.

Measured experiments meeting the criteria

Now 0 / goal 100

Independent sources (labs)

Now 0 / goal 5

Lower error than the baseline in lab-level validationNot yet

Validation with new, unseen experimentsNot yet