To use ChatGPT effectively, 5 principles you should know as a UI/UX designer

Instead of being replaced by AI, it’s better to collaborate with it.

The image is designed by Midjourney

Sharpen the tools before doing the job well.” ChatGPT is a large language model that can answer various questions through text input. Therefore, good questions are crucial for ChatGPT to perform well. By doing so, designers can not only save time but also spend more time on creative design and user research, allowing each other to do what they do best. So, how can we have effective communication with ChatGPT?

Principle 1: Don’t blindly trust ChatGPT results based on intuition

During the data generation process, AI models are like black boxes, and we do not understand how they organize and summarize information or their sources. This opacity may lead to unexplainable predictions, bias, or unfairness. Microsoft’s past research has shown that when AI systems fail, explanations may increase over-reliance and affect people’s judgments. Therefore, blindly trusting ChatGPT’s results may lead to some errors, such as inaccurate results, lack of subjectivity, unsuitability for some problems or domains, and inflexibility.

Principle 2: Precise and effective questioning strategies

Similar to human communication, we need to understand different ways of asking questions:

Open-ended questions

Using open-ended questions allows ChatGPT to answer freely, rather than being limited to a pre-set range of answers. For example: “What are your thoughts on the future development of user experience design?

Closed-ended question

Using closed-ended questions can provide specific options and allow ChatGPT to choose between them. For example: “Which research method do you think is more suitable for conducting user research on early-stage Vietnamese markets: desktop research, in-depth interviews, or field surveys?

Focus question

Using a focus question can concentrate discussion on a specific topic or issue. For example, “What are your thoughts or suggestions on using wearable devices in blockchain?

Open-ended probing question

Using open-ended probing questions can guide ChatGPT to further elaborate on their views and ideas. For example, “Could you explain your thoughts again?

Closed-ended probing question

Using closed-ended probing questions can restrict ChatGPT to specific options to gain a deeper understanding of their preferences and thoughts. For example, “Which one do you think is more suitable for application in the Chinese market, data verification, healthcare management, or identity authentication?

Principle 3: Using multiple languages to solve problems may not always be the most effective approach.

According to the official introduction from OpenAI, ChatGPT was trained on a large amount of English text. This means that ChatGPT’s language abilities may be stronger in English than in any language. However, using different languages can also lead to some issues.

For example, if ChatGPT is not familiar with a particular language, it may not fully understand the question and may therefore be unable to provide an accurate answer. In addition, using different languages can make communication more difficult and slow, especially when translation is involved. Therefore, during the interaction process, it is best to maintain the same language as much as possible, and avoid using ambiguous language, as a word or phrase may have multiple interpretations, especially when interpreted differently in different contexts.

Training Data for Various Languages

Principle 4: Correcting context can yield better results.

Providing appropriate background information, such as time, location, and related people, can help ChatGPT better understand the question. However, some limitations should also be noted. ChatGPT is a language model based on statistical learning, and its abilities depend on the quality and coverage of the training data. Therefore, if the context is very specific or rare, or if dialects, professional jargon, or informal language are used, ChatGPT’s effectiveness may be limited.

Furthermore, while ChatGPT can understand the semantics and grammatical rules of language, it may not necessarily understand the context and background of the conversation. For example, ChatGPT may not be able to determine whether a word’s meaning is positive or negative and may need to rely on the specific situation for judgement. Therefore, it is best to avoid overly abstract or hypothetical questions and to provide specific questions or requests as much as possible.

Errors in Contextual Switching

Principle 5: Start a new conversation at the right time

When asking questions to ChatGPT, it generates responses based on the question’s content and any relevant contextual information. It considers the question as part of the context and generates the best answer based on learned patterns and information in the context. Therefore, if the question is not meeting expectations, it may be helpful to start a new conversation. ChatGPT will begin its understanding and response process again, avoiding confusion with previous conversation content.

Restarting a new conversation flow to improve the quality of responses

Prompts for UI/UX designer's reference

User Research

• In what situations would users use our product or service?
• What are the main competitors and their primary features? What are their purposes?
• What are the significant differences between our product and the competition?
• What are the pros and cons of competitor designs, and how do they affect the user experience?
• List the main websites for a particular industry or product.
• List the most cited papers for a particular industry or product.
• Summarize research findings from a particular paper or article.
• Write an interview script.
• Write user testing tasks.
• Analyze results based on data.
• Based on feedback, what challenges or difficulties do users encounter when using the product or service?

Creating User Roles or Scenarios

• Create a user journey based on a particular scenario and pain point.
• Create a user journey based on a particular competitor.
• Write a story about a person using the product.

Copywriting

• Rewrite copy and provide more choices.
• Shorten strings and content descriptions.
• Multilingual translations.
• Establish word usage standards and guidelines.

Product Interaction Design

• List a product’s information architecture and main features.
• Summary of design guidelines for a particular product or professional field.
• Localization design points to consider for a particular country or region.
• Design details to consider for a particular demographic.

Visual Design

• Summarize the primary websites for inspiration.
• Establish design systems and guidelines.
• Create a color document.
• Design tool troubleshooting.
• Understanding design trends or element features.

Self-Growth

• Find bestselling books.
• Find books in the professional field.
• Organize book outlines.
• The most viewed article on a particular website.
• Cross-team design communication strategies.
• Design leadership and management methods.
• Explain complex terminology.