A shared memory.
A useful starting point.

The Knowledge Atlas is Darronhold Bridge's maintained internal knowledge system. Staff use it when developing programs, guides, workshops, and partner resources.

Put an idea to work.

Four selections from Kyle Hobson's Knowledge Atlas and AI Glossary collection. Each offers a different way to make information more useful: notice a risk, establish context, plan the work, or review the result.

Concept · Notice the risk

Automation bias

A recommendation can feel more trustworthy because it comes from an automated system. Automation bias is the tendency to rely on that recommendation even when other evidence should make us question it.

When it helps: When a fluent AI draft, summary, or recommendation is about to become someone else's reference material.
Try it: Choose one factual claim in an AI-generated guide. Find the underlying source and compare the wording, exceptions, and date. Record any disagreement before accepting the claim.
Watch out for: Knowing the name of the bias does not eliminate it. A check is a review practice, not a guarantee against error.

Collected in the Knowledge Atlas. Research reference: Raja Parasuraman and Dietrich H. Manzey (2010). The guide-review activity is a Darronhold teaching adaptation. Read the research abstract.

Method · Establish context

Facts & Expertise Primer

Before asking a model for an output, give it checked facts, the perspective it should work from, and a clear scope. Identify missing information instead of letting a plausible story fill the gaps.

When it helps: When drafting a resource that must reflect a particular service, audience, or situation.
Try it: For a workshop invitation, list the confirmed audience, duration, and activities on separate lines, with their sources. Mark unconfirmed details as unknown. Then request a draft limited to that information.
Watch out for: The model does not make a supplied fact true. Verify the inputs yourself and check that the draft has not introduced new claims.

Adapted from the Facts & Expertise Primer entry in Kyle Hobson's Knowledge Atlas / AI Glossary collection. The record does not identify an original external author; no origin claim is made. Open the collection entry.

Method · Make the work manageable

Goal Breakdown

Prompt/Question Decomposition

Turn a complex goal into smaller questions and tasks that can be checked. The result should make the next action clearer, not simply produce a longer to-do list.

When it helps: When a team knows the outcome it wants but cannot yet tell what to do first.
Try it: Break 'prepare a volunteer orientation' into four outputs: confirmed audience needs, a short agenda, a practice activity, and a reviewed guide. Name what would show that each output is ready, then choose the first unresolved question.
Watch out for: A detailed plan can still rest on a wrong assumption. Check dependencies and unknowns before carrying out the steps.

Adapted from Goal Breakdown (Prompt/Question Decomposition) in Kyle Hobson's Knowledge Atlas / AI Glossary collection. The record does not identify an original external author; no origin claim is made. Open the collection entry.

Framework · Review the result

RACCCA

Relevance · Accuracy · Completeness · Clarity · Coherence · Appropriateness

Use six lenses to review an AI-generated response. Instead of asking only whether a draft sounds good, look for the particular ways it succeeds or needs revision.

When it helps: When comparing drafts or preparing one for a specific audience.
Try it: Review a short response against the six criteria. Under each, note one observation and a concrete change. The worked example below shows this in practice.
Watch out for: Ratings are judgments, not objective measurements. Accuracy still needs evidence outside the response.

Andrew Maynard's RACCCA framework, developed for an Arizona State University prompt-engineering course; collected in Kyle Hobson's Knowledge Atlas. The example below is a new teaching adaptation. Read Maynard's framework.

Keep exploring: browse Kyle Hobson’s full framework database, with searchable methods, use cases, and prompt worksheets.

One method, in practice

From a vague invitation to a useful one.

An illustrative community-team example using RACCCA. This is an invented draft and revision, not a real engagement or measured outcome.

The information available

For this example only, the team has confirmed a 30-minute orientation for first-time volunteers who welcome visitors. It covers a welcome, a practice conversation, and time for questions. The schedule and venue are not yet confirmed; registration will open after those details are settled.

Before

Attend our comprehensive onboarding experience to become fully equipped for every situation. All relevant protocols will be covered and you'll leave ready to handle anything.

After review

New to volunteering? This 30-minute orientation introduces how our team welcomes visitors. We'll share a short welcome, practice one conversation together, and leave time for questions. The schedule and venue will be confirmed before registration opens.

The six review decisions

Relevance: Name the first-time volunteers and their orientation.
Accuracy: Remove the promise of readiness for every situation.
Completeness: Include the duration, activities, and missing logistics.
Clarity: Replace 'comprehensive onboarding' with everyday language.
Coherence: Put the purpose before the activities and next update.
Appropriateness: Use an inviting tone without promising expertise.

What still needs a person

Before sharing, a person still needs to confirm the logistics, accessibility arrangements, and whether the activities prepare volunteers for the situations they will actually meet. Better wording alone cannot establish that.

Try the same move

Try it with a paragraph of your own: which criterion reveals the most important change? Revise that part, then check whether the new wording adds any unsupported promises.

Keep the context
with the idea.

It preserves reusable knowledge together with the context needed to use it responsibly: what an idea means, where it came from, who it may help, and how it connects to other material.

Knowledge Stewardship maintains the shared collection. Community Learning, Resource Design, and Partner Enablement select and adapt material for a particular audience and setting.

What accompanies an entry

Classification: The kind of material: for example, a framework, method, or glossary term.
Audience: The people it is meant to support, such as facilitators, partner staff, or community learners.
Source: Where the idea or supporting information came from, so it can be checked.
Relationships: Connections to related terms, methods, or references that give an entry context.
Learning use: The purpose a resource can serve within a learning experience.
Review state: Review progress, distinct from suitability for a particular release.

Entries may carry classifications, audience labels, tags, source information, related terms, intended learning uses, and review states. These help staff find and evaluate material; inclusion in the Atlas does not mean an item is ready for every audience or use.

Different kinds of knowledge. Different uses.

Term or concept

Explains an idea and the language used to discuss it.

Framework or mental model

Offers a structure or perspective for thinking about a situation. It is not automatically a procedure.

Method or exercise

Describes an approach to trying something, practicing a skill, or guiding an activity.

Template or reference

Provides a reusable starting point or a source that supports further work.

Real ideas.
Illustrative situations.

Darronhold Bridge is a fictional organization created for a learning exercise. These concepts and methods draw on real educational material; the organizational examples are illustrative.

Site design, curation, and educational adaptation by Kyle Hobson. Original research and framework attribution are credited with the relevant entries.

These four adaptations are a small selection. The public framework library offers further reading; the fictional organization’s internal data format is not defined by this page.

Read the organization brief

A shared memory.A useful starting point.