Prompt-based methods have become increasingly popular among information extraction tasks (e.g., event argument extraction), especially in low-data scenarios. By formatting a fine-tuning task into a pre-training objective, prompt-based methods resolve the data scarce problem effectively. However, previous researches seldom investigate the discrepancy among different strategies on prompt formulation. In this work, we compare two kinds of prompts, name and ontology-based prompts, and reveal how ontology-based prompts exceed its counterpart in event argument extraction. Furthermore, we analyse the potential risk (e.g., biases) in ontology-based prompts via a causal view and propose a debiasing method using causal intervention. Experiments on three benchmarks demonstrate that modified by our debiasing method, the baseline model becomes more robust, with significant improvement in the resistance to adversarial attacks. \(^{1}\) Our code is available at this repository .