Description¶
Description is "effectively describing goals to prompt useful AI behaviors and outputs" (Anthropic's 4D Framework for AI fluency). In bioinformatics, it requires communicating goals, context, and biological assumptions to AI agents in the form of technical specifications that drive the expected behaviors and make outputs reproducible and interpretable.
Practice areas on this page
In the tables below, the Source column shows where each competency comes from: ISCB and ACTIVE competencies are drawn from existing frameworks; 4D-Bio competencies are original to this practice model.
1. Translating biological knowledge into precise AI instructions¶
| Category | Competency | Source |
|---|---|---|
| Bioscience domain knowledge | A3: Work at depth in at least one technical area aligned with the life sciences. | ISCB |
| Domain knowledge translation | Translate deep domain-specific biological mechanisms, technical terminology, and experimental variables into precise computational instructions for AI execution. | 4D-Bio |
| High-context prompting | Write structured, unambiguous prompts using advanced techniques to guide AI analysis. | 4D-Bio |
| Handle prompts as one more scientific resource that needs to be tested, documented, reported, reproduced, and cited. | 4D-Bio | |
| Adaptive instruction | Redirect AI systems dynamically when they drift from the expected analytical path or task boundaries. | 4D-Bio |
2. Structuring raw biological data for AI model context¶
| Category | Competency | Source |
|---|---|---|
| Metadata standardization | Apply highly descriptive system instructions that translate experimental designs and dataset annotations into clear parameters for AI ingestion. | 4D-Bio |
| Bioscience domain knowledge | B3: Prepare life science data for computational analysis. | ISCB |
3. Documenting and tracking AI interactions and software parameters for reproducibility¶
| Category | Competency | Source |
|---|---|---|
| Provenance logging | Record every prompt text, system configuration, model version, and seed value used to achieve a computational result. | 4D-Bio |
| Automated documentation | Generate exhaustive documentation that describes the functions, parameters, and expected outputs of AI-assisted code. | 4D-Bio |
| Version control | Track changes to scripts, prompts, and configurations regularly using version control systems like GitHub. | 4D-Bio |