Acesse a sua conta getAbstract para obter o resumo!

Auditing AI

Acesse a sua conta getAbstract para obter o resumo!

Auditing AI

MIT Press,

15 min. de leitura
7 Ideias Fundamentais
Áudio & Texto

Sobre o que é?

Artificial intelligence is reshaping decision-making; AI audits can help hold these systems accountable to human values.

  • 0.5×
  • 0.75×
  • 1×
  • 1.25×
  • 1.5×
  • 1.75×
1×

Editorial Rating

9

getAbstract Rating

  • Applicable
  • Eye Opening
  • Bold

Recommendation

AI is a growing facet of modern life, contributing to decision-making in a wide range of settings and professions, including medical care, policing, public policy, and HR. But AI systems are not unerring and must be carefully monitored for risks and biases. In this helpful guide, the AI auditing experts of the Marquand House Collective unpack why AI decision-making systems need to be assessed for accuracy and fairness. They explain how to conduct a thorough audit of these systems to help people in diverse fields decide when and how to use AI systems and drive positive societal change.

Summary

AI-driven decision-making is growing more common, bringing new benefits and risks.

Artificial intelligence (AI) systems are now a part of human life at just about every level. These systems are performing many tasks that most would agree serve human interests — identifying spam emails and calls on people’s smartphones, for example, or helping oncologists formulate treatment plans for cancer patients, and supporting researchers in developing drugs for a wide variety of illnesses. But society’s growing dependence on AI-driven decision-making, in one form or another, also creates risks, including the exacerbation of inequality and targeted discrimination.

To ensure AI systems do not lead people, organizations, and governments astray or negatively impact their safety and well-being, it’s vital to regularly audit them — that is, to assess their accuracy and reliability and whether biases are distorting the recommendations they provide.

Concerns about the accuracy and validity of automated systems that make value judgments are by no means new. In the early 2000s, an automated transplant prioritization...

About the Author

The Marquand House Collective comprises eleven experts in AI auditing spanning computing, law, policy, social science, and journalism: Marc Aidinoff, Lena Armstrong, Esha Bhandari, Ellery Roberts Biddle, Motahhare Eslami, Karrie Karahalios, Nate Matias, Danaé Metaxa, Alondra Nelson, Christian Sandvig, and Kristen Vaccaro.

Comment on this summary