| title | venue | abstract | transition | date |
|---|---|---|---|---|
Echical Challenges in Data Science |
HSBC Briefing |
None |
2020-06-01 |
\include{talk-macros.tex}
\include{_data-science/includes/evolved-relationship.md} \include{_ai/includes/embodiment-factors-short.md}
\include{_ml/includes/data-plus-model.md}
\include{_governance/includes/data-governance-toolkit.md}
\newslide{Data and Risk}
\define{\divoptions}{max-width:100vw; max-height:100vh} \define{\svgstyle}{height:50%} \define{\stubname}{new-flow-of-information}
\slides{\div{\includediagram{\diagramsDir/governance/risk-framework}{70%}{}{\svgstyle}}{\stubname}{\divoptions}}
\newslide{Data and Risk}
\define{\divoptions}{max-width:100vw; max-height:100vh} \define{\svgstyle}{height:50%} \define{\stubname}{new-flow-of-information}
\slides{\div{\includediagram{\diagramsDir/governance/risk-framework2}{70%}{}{\svgstyle}}{\stubname}{\divoptions}}
\newslide{Framework}
- Explainability
- Bias
- Data Sources
\newslide{Explainability}
- Purpose - define scope and why 'model + data'
- Rationale - what features drive outcome?
\newslide{Bias}
- Unrepresentative data
- Historic bias
- Proxy bias - model indirectly discriminates on protexted characteristics
\newslide{Data Sources}
- Consent - has permission been granted?
- Third party (Trojan ethics)
\thanks