On Making Lie Detectors

Our Attempt at shining a light into the "black box" of AI.

The Trusted Face began as an untruthfulness prototype developed for the 2020 election: a web-scale lie detector aimed at combatting disinformation. Since then, we’ve expanded our mission to not just combat untruthfulness in tech, but eliminate the lack of understanding that inevitably comes with its development. Machine Learning models are frequently described as “black box” models. Because the computer is, effectively, learning on its own, not even the researchers and computer scientists who have developed the algorithms can consistently explain why they often lead to incorrect or biased results. This creates a need for transparent, easy-to-perform experimentation on user-end models, with subsequent understanding of the context in which these models are being used and applied. Here, on our Blog page, you will find our human-centered, transparency-focused approach to artificial intelligence that is intended to help you, wherever you are in your AI journey, gain insight into the experiments, discussions, and insights that are shaping AI.