# PAIR-code/what-if-tool

Source code/webpage/demos for the What-If Tool

Repository: https://github.com/PAIR-code/what-if-tool
Canonical: https://ross.abutalabs.com/products/what-if-tool
Homepage: https://pair-code.github.io/what-if-tool
Language: HTML
License: Apache-2.0
License Family: permissive
Topics: ml-fairness, visualization, machine-learning, jupyterlab-extension, colaboratory, tensorboard
Last push: 2026-06-21T19:46:04+00:00

## Health v2 (maintenance only)
Score: 62/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 88, release rhythm 8, longevity 100
- inputs: {"age_days": 2917, "days_push": 73, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1007, forks 183 (observed 2026-08-28T04:03:12.460324+00:00)

## What it is
The What-If Tool (WIT) is a visual interface from Google's PAIR team for probing black-box classification and regression ML models without writing code. It runs inside TensorBoard or as a Jupyter/Colab notebook extension, letting users run inference, edit examples, and analyze model performance and fairness across data subsets.

## Use cases
- visually probe behavior of a trained ML model
- analyze model fairness across demographic subsets
- compare performance of multiple models side by side
- edit test examples and re-run inference to see what changes
- investigate feature importance and counterfactuals without coding
- explore model predictions in a Jupyter or Colab notebook

## When to choose
- you want no-code visual exploration of a trained classifier or regression model
- you need to audit model fairness and performance across data subsets
- you already work in TensorBoard, Jupyter, or Colab
- you want counterfactual what-if analysis of individual predictions

## When to avoid
- you need programmatic interpretability methods like SHAP or LIME in a pipeline
- your model is not servable via a Python function, TF Serving, or Cloud AI Platform
- you need real-time production model monitoring rather than interactive analysis
- you work outside the TensorFlow/Python notebook ecosystem

## Facets
- artifact type: plugin
- maturity: maintenance
- function: data-visualization, machine-learning, testing
- domain: machine-learning, data-visualization, artificial-intelligence
- platform: python, browser
- tags: model-interpretability, fairness, tensorboard-plugin, jupyterlab-extension, google-pair, no-code, model-analysis, web-server

## Member repositories
- PAIR-code/what-if-tool (main) score 62

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:12.460324+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:12:24.473765+00:00, confidence not recorded.
  - readme: https://github.com/PAIR-code/what-if-tool (fetched 2026-08-28T04:03:12.460324+00:00, sha 2cfffd3d718f)
  - homepage: https://pair-code.github.io/what-if-tool (fetched 2026-08-29T13:12:23.642771+00:00, sha 7df9bd2f79ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.