Themis AIBias Audit
Identify under/over-represented groups in your dataset. Understand the distribution of explicit and implicit features in your data.
At Themis AI we provide an end-to-end debiasing platform to companies that utilize AI in their decision-making pipeline. We help our customers identify bias in their data, mitigate it during training, and ultimately certify that a trained AI model is fair and ethical before deployment.
Identify under/over-represented groups in your dataset. Understand the distribution of explicit and implicit features in your data.
Mitigate algorithmic bias in your AI models. Optimize your data even on the challenging samples in your data.
Certify the fairness of your model on protected and underlying features. Ensure your AI model is ethical before deployment.
Co-founder & CEO
Co-founder & CTO
Co-founder & CSO
Co-founder & Chairwoman
"Interval, by The Engine, is a pilot program that offers MIT graduates, former MIT postdocs, and former MIT researchers a career path in entrepreneurship focused on pioneering Advanced Computing approaches to solve the world’s biggest and most urgent problems."
Alyssa Newcomb
"You can talk about bias as socially driven, the world needs to be a better place. But you can also talk about bias as bottle neck for extracting business value."
James Orme
"Facial recognition tasks have been shown to exhibit strong bias among certain demographics, such as women and minorities. In 2012, researchers showed that the face detection system used by the US police force was significantly less accurate at identifying 18-30-year-old women with darker skin."
Danielle Brown
"According to the paper, the team’s deep-learning algorithm can simultaneously learn desired tasks and underlying latent structure of the training data––meaning the algorithm can look at a dataset, learn what’s hidden inside it and automatically resample it to be more fair without using a programmer."
Kyle Wiggers
"Bias in algorithms is more common than you might think. An academic paper in 2012 showed that facial recognition systems from vendor Cognitec performed 5 to 10 percent worse on African Americans than on Caucasians, and researchers in 2011 found that models developed in China, Japan, and South Korea had difficulty distinguishing between Caucasians and East Asians."
Nancy Ordamn
"MIT CSAIL researchers recently reported the development of an algorithm that removes bias from the training data sets that teach it to categorize facial images. The technique may offer a promising way to eliminate racial bias in image recognition systems."