Collaborative training of fair machine learning models

A computer-implemented method comprising:

• obtaining, by a collaboratively trained machine learning model, an embedding vector, the collaboratively trained machine learning model having been collaboratively trained by first and second entities using federated learning;
• outputting, by the collaboratively trained machine learning model, an output vector, the output vector comprising a set of output vector element values;
• obtaining, by a first classification head, the set of output vector element values, the first classification head having been trained by the first entity to predict a value of a first attribute based on the set of output vector element values;
• outputting, by the first classification head, a predicted value of the first attribute based on the set of output vector element values, the predicted value of the first attribute being useable by the first entity to determine an accuracy of the first classification head in predicting values of the first attribute from output vector element values output by the collaboratively trained machine learning model;
• obtaining, by a second classification head, the set of output vector element values, the second classification head having been trained by the second entity to predict a value of a second attribute based on the set of output vector element values;
• outputting, by the second classification head, a predicted value of the second attribute based on the set of output vector element values, the predicted value of the second attribute being useable by the second entity to determine an accuracy of the second classification head in predicting values of the second attribute from output vector element values output by the collaboratively trained machine learning model;
• and
• updating the collaboratively trained machine learning model, based on at least the predicted value of the second attribute, to:
• increase the accuracy of the first classification head in predicting values of the first attribute from output vector element values output by the collaboratively trained machine learning model;
• and
• decrease the accuracy of the second classification head in predicting values of the second attribute from output vector element values output by the collaboratively trained machine learning model.