作者: Malin Eiband , Martin Zürn , Daniel Buschek
关键词:
摘要: We present an in-depth analysis of the impact multi-word suggestion choices from a neural language model on user behaviour regarding input and text composition in email writing. Our study for first time compares different numbers parallel suggestions, use by native non-native English writers, to explore trade-off "efficiency vs ideation", emerging recent literature. built editor prototype with (GPT-2), refined prestudy 30 people. In online (N=156), people composed emails four conditions (0/1/3/6 suggestions). results reveal (1) benefits ideation, costs efficiency, when suggesting multiple phrases; (2) that speakers benefit more suggestions; (3) further insights into patterns. discuss implications research, design interactive systems, vision supporting writers AI instead replacing them.