This paper introduces a new problem in multi-view learning called Reliable Conflictive Multi-View Learning (RCML), which requires models to provide decision results and their associated reliabilities for conflictive multi-view data. The proposed method, Evidential Conflictive Multi-View Learning (ECML), addresses the challenge of handling conflicting information across different views by learning view-specific evidence and aggregating conflicting opinions. ECML first learns view-specific evidence, which represents the support for each category from the data. It then constructs view-specific opinions consisting of belief masses and reliability. In the multi-view fusion stage, ECML proposes a conflictive opinion aggregation strategy that theoretically models the relationship between multi-view common and view-specific reliabilities. Experiments on six real-world datasets show that ECML outperforms existing methods in terms of accuracy, reliability, and robustness. The method ensures consistency between different views and accounts for potential conflicts between opinions from different views. Theoretical analysis confirms that the proposed aggregation strategy increases uncertainty for conflictive instances, making the model more reliable in decision-making. The code for ECML is available at https://github.com/jiajunsi/RCML.This paper introduces a new problem in multi-view learning called Reliable Conflictive Multi-View Learning (RCML), which requires models to provide decision results and their associated reliabilities for conflictive multi-view data. The proposed method, Evidential Conflictive Multi-View Learning (ECML), addresses the challenge of handling conflicting information across different views by learning view-specific evidence and aggregating conflicting opinions. ECML first learns view-specific evidence, which represents the support for each category from the data. It then constructs view-specific opinions consisting of belief masses and reliability. In the multi-view fusion stage, ECML proposes a conflictive opinion aggregation strategy that theoretically models the relationship between multi-view common and view-specific reliabilities. Experiments on six real-world datasets show that ECML outperforms existing methods in terms of accuracy, reliability, and robustness. The method ensures consistency between different views and accounts for potential conflicts between opinions from different views. Theoretical analysis confirms that the proposed aggregation strategy increases uncertainty for conflictive instances, making the model more reliable in decision-making. The code for ECML is available at https://github.com/jiajunsi/RCML.