We present an exploratory qualitative study to understand how writers interact with next-phrase suggestions. While there has been some quantitative research on the effects of suggestion systems on writing, there has been little qualitative work to understand how writers interact with suggestion systems and how it affects their writing process - specifically for a non-native but English writer. We conducted a study where amateur writers were asked to write two movie reviews each, one without suggestions and one with. We found writers interact with next-phrase suggestions in various complex ways - writers are able to abstract multiple parts of the suggestions and incorporate them within their writing - even when they disagree with the suggestion as a whole. The suggestion system also had various effects on the writing processes - contributing to different aspects of the writing process in unique ways. We propose a model of writer-suggestion interaction for writing with GPT-2 for a movie review writing task, followed by ways in which the model can be used for future research, along with outlining opportunities for research and design.
翻译:我们进行了一项探索性的定性研究,以了解作家如何与下一句话的建议互动。虽然对建议系统对写作的影响进行了一些定量研究,但对于作者如何与建议系统互动以及这种互动如何影响其写作过程,特别是对于一位非本地但英文的作家而言,却很少进行定性研究。我们进行了一项研究,要求业余作家分别撰写两份电影评论,一份没有建议,一份与英国同行。我们发现作家以各种复杂的方式与下一句话的建议互动,作者能够抽象地将建议的许多部分纳入他们的写作过程,即使他们不同意整个建议。建议系统也对写作过程产生了各种影响,以独特的方式促进了写作过程的不同方面。我们提出了与GPT-2的写作互动模式,以便与GPT-2的写作任务进行写作,随后是模型可用于未来研究的方式,同时概述了研究与设计的机会。