Promptteknik inom visuell kommunikation : en jämförande studie av hur multimodalitet påverkar teknikbaserad prompting

Sammanfattning

Generative AI has changed the conditions for visual communication. AI tools can quickly generate visual proposals, such as logos, color palettes and graphic elements. At the same time, AI introduces challenges when creative intentions must be translated into prompts that the AI can interpret and transform into relevant visual output. The study is based on a design assignment for United :Power, where AI generates proposals within three design areas: logo, color palette and graphics. The purpose is to compare how technique-based prompting methods, with and without multimodality, affect AI-generated design proposals. The study uses a small, exploratory and comparative approach with experimental elements. It includes three types of technique-based prompting: Direct Instruction Prompting, Persona-Based Prompting and Contextual Prompting. Each prompt is carried out both with and without multimodality. The AI-generated design proposals are assessed by five respondents with experience in visual communication, through surveys and semi-structured interviews. The proposals are evaluated based on originality, functionality and aesthetics. The study also examines the intention-to-prompt gap, meaning alignment between the intention of the prompt and the generated output. The results should be understood as exploratory and descriptive, as the study is based on five respondents and one evaluated AI-generated design proposal per prompt. In this sample, the observed differences between prompting with and without multimodality were limited. The results suggest that the visual reference material did not consistently affect the respondents’ assessments. This also suggests that multimodality, in some cases, may have created additional room for interpretation rather than reducing the intention-to-prompt gap. The study also suggests that AI was more successful in following concrete visual instructions than abstract design values, such as seriousness, credibility and subtle playfulness. The results emphasize the importance of clear prompts, deliberately designed reference material and critical human assessment when using generative AI in visual design processes.