Title: Language Style in Online Experiments
Advisors: Katharina Reinecke and Noah Smith
Getting high quality data in online experiments requires recruiting a large number of motivated participants and keeping them engaged throughout an experiment. One factor that impacts user behaviour and engagement in other online domains is language style, such as how formal a text is. This talk highlights two recent projects showing how the language style an online experiment uses to advertise and instruct participants in an online study can influence participant self-selection and engagement. The first project finds that the framing of a slogan, or what is emphasized when advertising a study, can significantly affect how many participants a study recruits and how motivated they are. The second project delves deeper into the experiment itself, showing that the formality of study instructions (e.g., "Please read the instructions carefully" vs. "Check out these instructions before you jump in!") can affect how attentive participants are during a study. We will then discuss future work on charactizing language styles automatically and identifiying their effect on user behavior in online communities.