Tools  |    |  April 1, 2020

RAI Design‌ ‌Assistant‌

Tool developed by the Responsible AI Institute (formerly AI Global) to assess work in development, to anticipate problems and future-proof AI systems. Access research and industry best practices to assess key challenges, including data rights and use, privacy, security, explainability, fairness, bias, and robustness. This‌ ‌is‌ ‌a‌ ‌virtual‌ ‌guide‌ ‌to‌ ‌help‌ ‌those‌ designing,‌ ‌developing,‌ ‌and‌ ‌implementing‌ ‌AI‌ ‌systems‌ ‌do‌ ‌so‌ ‌in‌ ‌a‌ ‌responsible‌ ‌way.‌

Committed‌ ‌to‌ ‌making‌ ‌responsible‌ ‌AI‌ ‌systems,‌ ‌we’ve‌ ‌done‌ ‌the‌ ‌hard‌ ‌work‌ ‌of‌ ‌deciphering‌ the‌ ‌best‌ ‌practices,‌ ‌policies,‌ ‌and‌ ‌principles‌ ‌and‌ ‌put‌ ‌them‌ ‌into‌ ‌a‌ ‌simple‌ ‌online‌ ‌survey.‌ With‌ ‌our‌ ‌esteemed‌ ‌community‌ ‌of‌ ‌subject‌ ‌matter‌ ‌experts‌ ‌ranging‌ ‌from‌ ‌engineers,‌ ‌to‌ ethicists,‌ ‌to‌ ‌policy‌ ‌makers,‌ ‌we‌ ‌have‌ ‌taken‌ ‌the‌ ‌most‌ ‌cited‌ ‌principles,‌ ‌whitepapers,‌ ‌and‌ policy‌ ‌documents‌ ‌published‌ ‌by‌ ‌academics,‌ ‌standards‌ ‌organizations,‌ ‌and‌ ‌companies‌ and‌ ‌translated‌ ‌them‌ ‌into‌ ‌comprehensive‌ ‌questions.‌

Based‌ ‌on‌ ‌our‌ ‌research‌ ‌and‌ ‌experience‌ ‌we‌ ‌have‌ ‌created‌ ‌a‌ ‌comprehensive‌ ‌evaluation‌ looking‌ ‌at‌ ‌the‌ ‌following‌ ‌dimensions‌ ‌of‌ ‌a‌ ‌trusted‌ ‌AI‌ ‌program:‌

  • Organization Maturity
  • Accountability
  • Data
  • Fairness
  • Interpretability
  • Robustness

Our‌ ‌hope‌ ‌is‌ ‌that‌ ‌you‌ ‌will‌ ‌work‌ ‌with‌ ‌your‌ ‌colleagues‌ ‌who‌ ‌are‌ ‌responsible‌ ‌for‌ ‌different‌ aspects‌ ‌of‌ ‌your‌ ‌business‌ ‌to‌ ‌fill‌ ‌out‌ ‌the‌ ‌Design‌ ‌Assistant.‌ ‌Whether‌ ‌you‌ ‌are‌ ‌just‌ ‌thinking‌ about‌ ‌how‌ ‌to‌ ‌integrate‌ ‌AI‌ ‌tools‌ ‌into‌ ‌your‌ ‌business,‌ ‌or‌ ‌you‌ ‌have‌ ‌already‌ ‌deployed‌ several‌ ‌models,‌ ‌this‌ ‌tool‌ ‌is‌ ‌for‌ ‌you.‌ ‌We‌ ‌do‌ ‌think‌ ‌that‌ ‌these‌ ‌questions‌ ‌are‌ ‌best‌ ‌to‌ ‌think‌ about‌ ‌at‌ ‌the‌ ‌start‌ ‌of‌ ‌your‌ ‌project,‌ ‌however,‌ ‌we‌ ‌do‌ ‌think‌ ‌that‌ ‌the‌ ‌Design‌ ‌Assistant‌ ‌can‌ ‌be‌ used‌ ‌throughout‌ ‌the‌ ‌lifecycle‌ ‌of‌ ‌your‌ ‌project!‌

The responsible AI Design Assistant is an online assessment to help AI producers anticipate potential ethics issues in their AI system. This tool brings together research and industry leading  practices to help designers, developers, and product owners  keep in mind key AI risks including aspects of accountability and compliance, data rights and use, explainability, bias and fairness, and robustness.

The responsible AI Design Assistant constitutes the following:

  • A User Guide: Definitions and additional information which will provide you with the support needed to work through the Responsible AI Design Assistant self assessment
  • Responsible AI Design Assistant: Tool with questions to assess accountability, explainability and interpretability, data quality, bias and fairness and robustness pertaining to an AI system