Recent conversational recommender systems (CRSs) have achieved considerable success on addressing the cold-start problem. While they utilize conversational key-terms to efficiently elicit user preferences, most of them, however, neglect that key-terms can also introduce biases. Systems learning key-term-level user preferences may make a biased item recommendation based on an overrated key-term instead of the item itself. As key-term conversation is a crucial part of CRSs, it is important to properly handle such bias resulting from the item-key-term relationship. While many debiasing methods have been proposed for traditional recommender systems, most of them focus on items or item groups re-ranking or re-weighting strategies such as calibration and propensity score, which are not designed to model the relation between item and key-term user preference. There is also no effective way for traditional debiasing methods to measure potentially useful biases through conversational key-terms to enhance the recommendation performance. In this paper, we develop a deconfounded CRS, which enables the user to provide both item and key-term feedback in each round such that we can promisingly capture more accurate relation between key-term-level and item-level user preference to alleviate the bias. To better model the relations and understand such bias, we view CRSs from a causal perspective and introduce a novel structural causal model (SCM) that identifies the confounding effect of key-term-level user preference. Inspired by our causal view, we devise an online backdoor adjustment approximation to alleviate the confounding effect when making item recommendations. Consider that not all biases are harmful, we utilize the useful bias and propose DecUCB, which leverages conversational key-term feedback to regulate the influence of backdoor adjustment adaptively in a personalized fashion. Extensive experiments on real-world datasets demonstrate the advantages of our proposed method in both recommendation performance and bias mitigation.