This is the dataset for ChatbotManip: A Dataset to Facilitate Evaluation and Oversight of Manipulative Chatbot Behaviour
NOTE: this data is NOT for commercial use
This data contains 'highlighted_text' in survey_responses, which were the conversations that were highlighted by annotators to be generally manipulative. This is not written about in the paper, but might be useful for further analysis