Gender deanonymization through machine learning - risks and implications.

From IFF Wiki
Jump to: navigation, search
Gender deanonymization through machine learning - risks and implications.
Presenter(s) Suchana Seth
Title(s) Gender deanonymization through machine learning - risks and implications.
Organization(s) Data & Society, Ford-Mozilla Open Web Fellow
Project(s) https://datasociety.net/people/seth-suchana/
Country(ies) India, USA
Social media https://twitter.com/suchanaseth
2017 theme Tools & Technology


Date: Wednesday, March 8, 2017

Time: 14.30 hrs

The goal of this session is to spark a conversation among participants about the potential risks of gender deanonymization made possible by the latest advances in machine learning and data science.

The speaker will share examples and unpack the ideas behind algorithms that can predict gender using a variety of signals, as well as outline some of the risks to user privacy from such algorithms. The speaker will then illustrate scenarios where such gender deanonymization might result in disproportionate impact on vulnerable groups.

Participants will be asked to think about the risks of algorithmic gender deanonymization in the context of their life, work or activism, and share ideas about how they might begin to mitigate some of these risks.

We will also explore ideas around how we can make algorithms accountable for such privacy risks, and how users can audit the algorithms that impact them.

Format Conversation
Target Groups Activists, data scientists, tech policy advocates, folks interested in machine learning & data anonymization
Length 1 hour
Skill Level Novice
Language English


Session Outputs

Next Steps

Additional Notes

Relevant Resources

Contributors