Principles and guidelines

When used thoughtfully, AI can enhance user experiences through personalized interactions, increased efficiency, and innovative designs. PatternFly AI provides resources that can help you integrate AI into your design process, balancing a consideration for its potential benefits and unintended consequences.

Red Hat policies

When using PatternFly to design Red Hat products, you must adhere to AI-related policies that Red Hat has previously outlined:

Gain approval before using AI technology for business related to Red Hat.
Gain approval before using certain information as input for AI technology.
Review, test, and validate generative AI model output.
Always consider data privacy when entering company or personal information into AI resources, and ensure compliance with all company data protection policies and rules around AI usage.

View policy details (requires Red Hat login)

Core principles

There are 5 core principles of PatternFly AI: accountability, explainability, transparency, fairness, and human-centeredness. These principles create an ethics-first framework for AI use, and any AI system built with PatternFly should adhere to all 5.

All people involved in any step of creating AI are accountable for considering its impact. There should be clearly defined roles for design, development, and deployment. Decisions and processes should be well-documented.

AI systems should be easy to explain and comprehend. Humans should be able to easily perceive, recognize, and understand their decision-making processes. Imperceptible AI is not ethical.

All processes, decisions, and practices involved in AI systems should be open and transparent—not hidden. Users should be able to understand who is making decisions and how these decisions are made.

AI systems that are fair should be intentionally designed to prioritize and promote inclusion. They should focus accessibility for all humans and should minimize—not amplify—bias.

At their core, AI systems must prioritize a human-centered approach, focusing on addressing real needs and empowering humans. They should emphasize accessibility, transparency, user autonomy, and privacy. To build trust and promote understanding, these systems should provide users with intuitive interfaces and clear communication of intentions and capabilities.

Ethical design checklist

When working on an AI system, you should consciously check that you're in alignment with the core principles of PatternFly AI. While this is an area that will continue to evolve with the rest of the industry, the following checklists outline some of the key areas that you should consider for each principle.

Accountability

Key area	Rule	Status
Policies	Company AI policies are readily accessible to all team members.	Compliant
Legal compliance	All necessary laws, regulations, and ethical guidelines are followed throughout the development process. AI does not enable illegal, unethical, or contract-breaking activities.	Compliant
Practices	AI does not answer unsafe questions or access unsecure data.	Compliant

Explainability

Key area	Rule	Status
Outcomes	There are clear explanations available that describe how AI conclusions are reached.	Compliant
Citations	Any related citations are provided to users.	Compliant
Context	To support troubleshooting, AI gives context to Red Hatters who review its interactions.	Compliant

Transparency

Key area	Rule	Status
Documentation	Design processes and decisions are well documented.	Compliant
Data usage	Informed consent is obtained to collect and use data. The ways that user data is collected, stored, and used are openly shared. AI is clear about the data that it records.	Compliant
Confidence	AI shares when it has low confidence in its response.	Compliant
Limitations	AI shares when it believes that it can’t fulfill a request.	Compliant

Fairness

Key area	Rule	Status
Bias	Potential biases are identified, reduced, and actively studied.	Compliant
Inclusion	Designs are inclusive and accommodating of various user demographics.	Compliant
Equal access	Access to AI technologies is available and beneficial to as many users and communities as possible.	Compliant

Human-centeredness

Key area	Rule	Status
Value and need	AI is aligned with user needs and values and will be continuously refined based on user feedback and ethical considerations.	Compliant
Communication	AI has a predictable tone and voice. It can handle emotional responses from users gracefully.	Compliant
Cultural sensitivity	Cultural differences are considered and respected.	Compliant
Data rights and control	Users have control over their data, including the ability to access, modify, and delete their information. AI does not act on behalf of users without explicit permission and clear opportunities for permission withdrawal.	Compliant
Optional	There is an obvious and simple way for users to opt out of using AI.	Compliant
Privacy	Personally identifiable information is protected and used responsibly.	Compliant

Design guidelines

When designing, developing, and using AI, consider the following ethical and best-practice guidelines.

1. Determine if AI adds value

Not all uses of AI are good for your UX strategy. As much as possible, conduct research to identify real user needs that AI features could help solve

Some of the more common problems that AI might be able to help solve include:

Increasing users' productivity and efficiency.
Personalizing user experience to make engagements more personal and relevant.
Making design processes more sustainable.

When to use AI

Depending on your users' needs, value-adding features could include:

AI-driven search, to tailor search results to a user's unique needs.
AI that helps streamline onboarding, data entry, or routine job tasks.
AI that makes product recommendations based on a user's history.

When not to use AI

Do not add AI features simply because they are new, trendy, or fun. They need to matter to the user.

2. Enhance—don't replace—human abilities

AI is best when it enhances human abilities, not when it's used to replace humans. It cannot exist in a silo—humans help bring the value of AI to life.

To ensure that the design of AI systems is human-centered, follow these practices:

Nurture collaboration and cross-team alignment.
Welcome multiple perspectives to encourage creativity and help mitigate bias.
Check AI output for accuracy and identify areas where meaning is lost, language isn't inclusive, or information isn't true. Ask peers to review your AI-generated deliverables to help fact-check and catch mistakes.

3. Be transparent with your users

As one of our core pillars, transparency is essential for ethical design with AI.

To help people understand and trust AI features:

Tell users when AI is being used.
Make its capabilities and limitations clear to set appropriate expectations.
Explain how AI makes decisions.
Keep users in control and let them decide how they interact with AI.
Be clear and honest when AI fails or hallucinates.

4. Be prepared for something to go wrong

Errors and failure are inevitable when working with AI, so it is essential that you are prepared to handle undesired outcomes. You should understand the risk involved in AI and the impact that an error may have.

To create a plan for issues, start by following these guidelines:

When AI fails, be explicit about errors and let users regain control as they want.
Provide easy access to human support.

Edit this page on GitHub