AI prompt writing techniques

This project asks whether AI prompt writing skills could assist freelance translators in day-to-day tasks, and especially terminology management, time management and productivity. A particular challenge associated with translating Spanish medical texts is the use of abbreviations, which should therefore provide a useful case study. 25 clinical trials abbreviations were chosen and translated in advance. Four prompts were pre-written emulating different levels of prompt writing training, ranging from no training through to an expert prompt writer. A mixed-methods approach was used; AI responses were scored as either correct or incorrect against the human translation and the AI output was examined in thematic clusters. Key themes included AI’s need for context, and freelance translators’ need to retain their critical thinking skills, subject matter expertise and professional translation knowledge regardless of the prompt. The main finding is that AI accuracy did not increase with prompt complexity as expected. We therefore conclude that translators do not need to undertake extensive prompt writing training. Findings have implications for translator training: Rather than training prompt writing itself, courses should focus on examining the accuracy of claims about ways in which AI tools should be used.