Background: Social media platforms have introduced new opportunities for supporting family caregivers of persons with Alzheimer’s disease and related dementias (ADRD). Existing methods for exploring online information seeking and sharing (i.e., information exchange) involve examining online posts via manual analysis by human experts or fully automated data-driven exploration through text classification. Both methods have limitations. Design: In this paper, we propose an innovative expert–machine co-development (EMC) process that enables rich interactions and co-learning between human experts and automatic algorithms. Results: By applying the EMC in analyzing ADRD caregivers’ online behaviors, we illustrate steps required by the EMC, and demonstrate its effectiveness in enhancing human experts’ representations of ADRD caregivers’ online information exchange and developing more accurate automatic classification models for ADRD caregivers’ information exchange.