Towards a Theory of Longitudinal Trust Calibration in Human–Robot Teams