Methods, systems, and computer program products are provided. Details of a calendar item may be extracted. A conflict involving the calendar item and a second calendar item may be detected. Plural votes related to an importance of the calendar item may be received on behalf of a trusted crowd. When a vote is not received on behalf of a member of the trusted crowd within a certain time period, a prediction may be made regarding how the member of the trusted crowd would vote based on using machine learning to analyze a voting history of the member of the trusted crowd with respect to calendar items related to a subject of the calendar item. The predicted vote may be submitted on behalf of the member. One of the calendar item and the second calendar item may be altered based on a tally of the plural votes to resolve the conflict.