In other words, the temperature value adjusts the style of a generative AI model’s outputs. For example, a high temperature may be suitable for songwriting or brainstorming, while a low temperature may be preferable for drafting formal documents. Put differently, increasing the temperature raises the likelihood of hallucinations, while lowering it increases the chance of repeating patterns from the training data. Prompt A prompt is the input data or instruction given to a generative AI model. In a text generation model, prompts take the form of sentences such as “Summarize this document.” In an image generation model, a description like “an illustration of a golden sun against a blue ocean background” serves as the prompt. In multimodal systems, prompts may combine text, images, and other forms of input to create more complex instructions. Prompt engineering refers to the practice of designing prompts to obtain desired outputs. For example, the prompts “Explain this briefly” and “Explain this with analogies and examples that an elementary school student can understand” will produce entirely different answers to the same underlying question. Prompt engineering encompasses strategies such as providing detailed instructions, assigning roles, including examples, or specifying output formats in order to use generative AI effectively. Generative AI Guide for Civil Society