Computer science professor Fernanda Viégas believes “dashboards” that disclose AI models’ biases can help laypeople control AI.