Aim for Just Enough

Overview

Too much information becomes overwhelming, too little information can confuse readers further. As the creator of a Data Card, you will need to curate and prioritize the information in it. A good transparency artifact will provide just enough context for readers to walk away with a clear understanding, and if not, they’ll know where to go from here.

You want to provide information that makes the dataset both easy to understand and to use, but as dataset complexity increases, so does the density of information and explanations. Readers of any expertise level can experience information overload, so it’s important to present the “right” information. This is a combination of three things – the kind of information, the quantity of information, and the detail in it. Descriptions that are supported with enough background and context can be satisfying and sufficient for a majority of decisions that readers make. Linking to additional documents can help readers find more information without overwhelming them. However, this comes at a cost – linking to poor or very large documents can derail readers from their immediate task at hand, and can complicate their understanding of the dataset.

Key Takeaways

• Keep in mind the quantity, quality, and content of documents being linked and summarized.
• Answers are made of summaries, insights and context. Summarize everything without detailing everything.

Actions

1. Structure and context. If you have a structured way of describing your system that has worked in the past, leverage it to help readers create working mental models of your dataset that they can default to in decision making.
2. Merging documentation. If you’re copying content over, make sure you’re not missing prior context. Scan the source material for any context that should be included in the summary or elsewhere in the Data Card.
3. Summarize links and documents. Scaffold links in summaries that provide a flavor of the linked artifact. This helps readers find the right information and maintain context. In your Data Card, emphasize crucial information that you don’t want readers to miss from the linked documents.
4. Don’t assume background knowledge. Assume that your readers are new to the material you present in the Data Card. Use captions for tables, visualizations, and images.

Considerations

• Solicit input from subject matter & domain experts about any background information or analyses that might be required to interpret answers in their contexts.
• Solicit input from multiple stakeholders who were involved in the project to ensure your Data Card and supplementary documentation is complete and accurate.
• Run user studies to evaluate how relevant your answers are to your readers. Solicit input from people who represent your agents or readers and check if your Data Card has appropriate levels of detail, sufficient background, the correct technical jargon, length, language, or format.