We thank the reviewers and the organizing committee for their constructive comments on our paper. In the following, we focus on the ethical concerns raised, but we also accounted for the other comments in the revised version. To account for the ethical concerns, we added an extra paragraph to the end of the introduction (emphasized with a paragraph heading) in order to highlight the potential risks of a change in meaning due to agency/power rewriting upfront. In more detail, we aimed to address the meta-review and the following comments by the reviewers: - "I would also like to draw more attention to an ethical problem which the authors discuss in the ethics section but not within the body of the paper. As they say, using counterfactual data for bias mitigation carries risks if the meanings of the sentences changes." (Reviewer #1) - "Changing the connotation of a sentence as done here, however, always changes the meaning to at least some extent. For instance, consider the #metoo articles discussed by Field et al 2019, in which we find statements like "The first woman to speak out was Leeann Tweeden who said that Franken forcibly kissed her," labeled as low agency and low power. It would be extremely offensive to rephrase this statement to emphasize Tweeden's agency while retaining the basic event structure, for example "Leann Tweeden said that she forced/opted for Franken to kiss her." (Reviewer #1) - "Are there cases when the changing agency/power could be bad? Without a clear articulation of the potential harms, it is not clear what the answer to these questions will be." (Reviewer #3) We found the following example by the first reviewer especially helpful to understand the potential harms and adopted it in the paper for better illustration (in the "Ethical Consideration" part of the introduction; with a mention in the acknowledgments): - "A well-meaning engineer who used this system for bias mitigation might train a language model on a text corpus in which slaves 'chose' rather than 'being forced' to work in cotton fields, or litigants like Thurgood Marshall 'demanded' rather than 'arguing for' social reforms. [...] Casting victims as complicit or even controlling abuses against them is a widespread trope in apologism for atrocities (e.g. the narrative that Black people were active participants in slavery, that Jewish conspirators masterminded the Holocaust, etc.)." (Reviewer #1) Besides, please recall that an extended discussion of ethical concerns at the end of the paper (Section 9).