Gender bias may emerge from an unequal representation of agency and power, for example, by portraying women frequently as passive and powerless (“She accepted her future”) and men as proactive and powerful (“He chose his future”). When language models learn from respective texts, they may reproduce or even amplify the bias. An effective way to mitigate bias is to generate counterfactual sentences with opposite agency and power to the training. Recent work targeted agency-specific verbs from a lexicon to this end. We argue that this is insufficient, due to the interaction of agency and power and their dependence on context. In this paper, we thus develop a new rewriting model that identifies verbs with the desired agency and power in the context of the given sentence. The verbs’ probability is then boosted to encourage the model to rewrite both connotations jointly. According to automatic metrics, our model effectively controls for power while being competitive in agency to the state of the art. In our main evaluation, human annotators favored its counterfactuals in terms of both connotations, also deeming its meaning preservation better.

An argument is a constellation of premises reasoning towards a certain conclusion. The automatic generation of conclusions is becoming a very prominent task, raising the need for automatic measures to assess the quality of these generated conclusions. The SharedTask at the 9th Workshop on Argument Mining proposes a new task to assess the novelty and validity of a conclusion given a set of premises. In this paper, we present a multitask learning approach that transfers the knowledge learned from the natural language inference task to the tasks at hand. Evaluation results indicate the importance of both knowledge transfer and joint learning, placing our approach in the fifth place with strong results compared to baselines.