Differential privacy in Workplace Analytics
Workplace Analytics is serious about protecting individual privacy. Privacy can always be guaranteed if no information is revealed, which is not very useful. Similarly, making all information available can lead to high-fidelity metrics that compromise individual privacy.
Differential privacy offers a balance between providing useful information and protecting individual privacy. Workplace Analytics uses methods from world-class researchers to apply differential privacy. By introducing slight variations to the data, it protects privacy while simultaneously maintaining accuracy. The methods are more sophisticated than this simple description, with numerous options toward balancing fidelity and privacy. For more details, see Differential Privacy for Everyone.
The first application of differential privacy in Workplace Analytics is within Insights for people managers. These insights enable managers to understand how the people in their team are doing and to learn how to drive change by using aggregated collaboration data.
No matter how the metrics are presented or used in Workplace Analytics, no individual activity or metric can ever be discerned. The individual activity can never be established with certainty, and no manager or team-leader can conclude with confidence anything specific about any individual.
See Differential privacy to learn more about how Microsoft AI is helping to define and use it.