FairLens unveiled: A world first in discovering and measuring data bias | Hacker News Reader

FairLens unveiled: A world first in discovering and measuring data bias

synthesized.io

4 pointstimreidmedia3 comments

Developers and data scientists invited to use open-source FairLens to focus on data bias solutions

August 9th, 2021

PRESS RELEASE

Synthesized, the leading all-in-one DataOps platform, has today released the world’s first data centric open-source library for identifying data bias and driving fairness in decision-making.

FairLens is an open-source Python library which allows data scientists to automatically discover hidden biases and measure fairness in data.

FairLens is instantly accessible and available in GitHub where Python developers can integrate FairLens into their workflow, collaborate or help develop new features.

Denis Borovikov, co-founder and chief technology officer at Synthesized, said: “The goal of FairLens is to enable data practitioners to gain a deeper understanding of their data, and to help ensure fair and ethical use of data in analysis and data science tasks.

“FairLens is a mathematical framework that we have developed as a starting point for discovering data bias. Synthesized is now calling on developers to get involved to expand ways in which bias and fairness can be tackled and potentially enhance and improve FairLens”.

Nicolai Baldin, co-founder and chief executive of Synthesized said: “Many data science models rely on biased and skewed datasets. What we have created, with FairLens, is a way to highlight the data bias so it’s easily discovered and visualised. While data bias is still a taboo subject for many companies and industries, what FairLens enables is a behind-the-scenes discovery of data bias, which can then be mitigated.”

“With the help of the developer and data science communities, and our machine learning technology, we can build and enhance FairLens collectively, driving adoption across organisations and communities and by doing so, we hope to make a valuable impact on society.”

Synthesized’s package provides multiple metrics to measure biases across a range of protected characteristics such as age, race, sex and other groupings. With limited, poor-quality or skewed datasets; data-driven applications often fail to achieve their intended purpose as they are inherently biased.

FairLens can be tried in Synthesized’s SDK Colab alongside other platform functionalities. It comes with core features which allow users to:

Measure bias - FairLens can be used to measure the extent and significance of biases in datasets using a wide range of metrics Identify sensitive attributes - Data Scientists can automatically identify and flag sensitive columns and hidden correlations between columns to protect sensitive attributes Visualise bias - For example visualize the distribution of a variable with respect to different sensitive demographics, or a correlation heat-map. Score fairness - Data Scientists can highlight hidden biases and correlations within a dataset by selecting a target variable

ENDS

For further information or interviews, please contact:

Tim Reid Media, PR and communications consultant 07720 414205

Email: tim@timreidmedia.com

I'm confused. Other than alerting to small sample size, how can this tool measure bias without knowing what the true distribution is?

For example, given a database of violent crime convictions, would it be marked as biased because men make up over 70% of offenders, instead of an "unbiased" 50%?

This initial version assumes that the entire dataset is a fair representation of the underlying phenomena, i.e. the "true" distribution can be estimated from data already. As the package develops and users of the package have more knowledge about different true distributions, the tool should be able to operate without this assumption. Hence aggregating knowledge about different distributions to eventually remove the assumption is part of the roadmap.

Reply on news.ycombinator.com