Exploring Language for Enhanced AI Interaction
Introduction
AI systems offer a unique opportunity for us to deepen our understanding of communication. Having studied AI in the early 1980s, I decided to revisit some theories from that time to see how they might help in extracting useful information from AI Systems. The first theories I explored were:
Minsky's Frame System Theory.
Role Theory.
Halliday's Functional Grammar.
Speech Acts.
With the system prompt shown in Appendix 1, Language Frames can be generated for any situation. The output of an AI System, customized with this system prompt, is shown in Appendix 2. The output was created from this input:
"Please create a Language Frame for a visit to the dentist in France with useful phrases and vocabulary grouped according to roles, speech acts, and functional grammar."
Not happy with just three theories, I decided to include more, in fact, 14 more. The modified "system prompt" for customizing an AI system is shown in Appendix 3. The prompt looks complicated, but fortunately AI is OK with that. With this modified system prompt, some quite informative conversations can be had about designing prompts for AI Systems.
This article is very short because I recommend inputting the system prompts from the appendices into an AI System and then asking questions about the method and, of course, trying it out.
Introduction to the Theories
Each of the theories briefly described below contributes uniquely to the topic of writing effective prompts. For instance, Distributed Cognition Theory mirrors the way communication has developed with the advent of project management techniques such as Agile and the associated software.
1. Role Theory (https://en.wikipedia.org/wiki/Role_theory)
Role Theory examines how individuals behave in different social roles, which come with specific expectations for behaviour and responses. This theory helps in defining clear roles in prompts to guide interactions and set expectations.
2. Minsky's Frame System Theory (https://en.wikipedia.org/wiki/Frame_(artificial_intelligence))
Minsky's Frame System Theory provides a structured way to understand scenarios by detailing the setting, roles, goals, actions, and relationships. This helps create contextually rich prompts that are well-organized and comprehensive.
3. Speech Acts Theory (https://en.wikipedia.org/wiki/Speech_act)
Speech Acts Theory explores how utterances function as actions (e.g., requesting, commanding, apologizing) rather than merely conveying information. It helps in specifying the communicative intent of prompts to ensure clear and purposeful communication.
4. Halliday's Functional Systemic Grammar (FSG) (https://en.wikipedia.org/wiki/Systemic_functional_grammar)
Halliday's FSG emphasizes the functional aspects of language, focusing on ideational (content), interpersonal (social relations), and textual (structure) functions. This theory ensures that language in prompts is diverse and contextually appropriate.
5. Activity Theory (https://en.wikipedia.org/wiki/Activity_theory)
Activity Theory examines the broader context of human activities and the tools used. It considers how people interact with their environment and tools, helping design prompts that are relevant to the user's activities and context.
6. Distributed Cognition Theory (https://en.wikipedia.org/wiki/Distributed_cognition)
Distributed Cognition Theory looks at how cognitive processes are shared across individuals, artefacts, and the environment. It encourages designing prompts that consider these shared cognitive processes to enhance collaboration and understanding.
7. Situated Learning Theory (https://en.wikipedia.org/wiki/Situated_learning)
Situated Learning Theory posits that learning occurs best in context. It emphasizes real-world contexts to make interactions more relatable and effective, helping users learn more naturally and effectively.
8. Communication Accommodation Theory (https://en.wikipedia.org/wiki/Communication_accommodation_theory)
Communication Accommodation Theory examines how individuals adjust their communication styles to accommodate others. This theory guides the adaptation of prompts to different audiences to improve understanding and effectiveness.
9. Politeness Theory (https://en.wikipedia.org/wiki/Politeness_theory)
Politeness Theory explores how individuals use language to show respect, avoid offence, and maintain social harmony. It ensures that prompts consider politeness and social appropriateness to foster positive interactions.
10. Cognitive Load Theory (https://en.wikipedia.org/wiki/Cognitive_load)
Cognitive Load Theory focuses on managing mental effort to avoid overload and ensure clarity. It helps design prompts that are easy to understand and process, preventing cognitive overload.
11. Flow Theory (https://en.wikipedia.org/wiki/Flow_(psychology))
Flow Theory studies how individuals become fully immersed in activities, leading to optimal engagement and enjoyment. It helps create prompts that facilitate engaging and immersive experiences for users.
12. User-centred Design (UCD) Principles (https://en.wikipedia.org/wiki/User-centered_design)
UCD principles emphasize designing with the user's needs, contexts, and experiences in mind. This approach ensures that prompts are user-friendly, relevant, and satisfying, enhancing overall usability.
13. Media Richness Theory (https://en.wikipedia.org/wiki/Media_richness_theory)
Media Richness Theory evaluates the effectiveness of different communication channels. It guides the selection of the most appropriate medium for a given message to ensure effective communication.
14. Grice's Cooperative Principle and Maxims (https://en.wikipedia.org/wiki/Cooperative_principle)
Grice's Cooperative Principle and Maxims advocate for clear, relevant, informative, and appropriately detailed communication. These principles ensure that prompts are cooperative and effective in conveying information.
Resulting Guidelines
The extended system prompt suggests the following list of guidelines for writing effective prompts for AI systems, based on the Enhanced Integrated Prompt Design Framework (EIPDF):
Understand the User's Goal: Identify the purpose behind the user's need for the prompt.
Define Roles Clearly: Specify the roles involved to set expectations for behaviour and responses.
Frame the Context: Provide detailed scenarios to guide understanding and responses.
Specify Intent: Clarify the communicative action (e.g., request, command) to direct responses.
Ensure Functional Diversity: Use diverse and contextually appropriate language, focusing on ideational, interpersonal, and textual functions.
Consider Activity Context: Reflect the broader context of user activities and tools.
Encourage Shared Cognition: Design prompts that consider shared cognitive processes.
Reflect Real-World Contexts: Use real-world contexts to make interactions relatable and effective.
Adapt Communication Styles: Adjust communication strategies for different audiences.
Maintain Politeness: Ensure politeness and social appropriateness in prompts.
Manage Cognitive Load: Design prompts to manage mental effort and avoid overload.
Facilitate Engagement: Create immersive and engaging experiences.
Align with User Needs: Apply user-centred design principles to enhance usability and satisfaction.
Choose Effective Mediums: Select appropriate communication channels and methods.
Ensure Clarity and Cooperation: Follow Grice's Cooperative Principle and Maxims for clear and effective communication.
With the advent of multi-modal AI Systems and their ability to retrieve and store information in the software systems we use, the above list is an interesting starting point for thinking about prompt design.
Appendix I
You are a supportive and knowledgeable mentor in the English language. You are also an expert in:
Minsky's Frame System Theory.
Role Theory.
Halliday's Functional Grammar.
Speech Acts.
Provide detailed, patient explanations and guidance on language usage, grammar, vocabulary, and communication skills within the context of the above theories. Tailor your responses to the user's level, whether they are a native or non-native speaker. Offer encouragement and constructive feedback to help build confidence and improve proficiency. Where appropriate, include examples and analogies to clarify points, and suggest exercises or activities to reinforce learning. Be attentive to the user’s questions and responsive in a manner that promotes a positive and engaging experience.
Example user prompts are:
“Please create a Language Frame for a visit to the dentist in France with useful phrases and vocabulary grouped according to roles, speech acts, and functional grammar.”
“Please create a Language Frame for going skiing in Austria, with useful phrases and vocabulary grouped according to speech acts and roles.”
Appendix II
Context: Visit to the Dentist
Roles and Associated Vocabulary
Patient (You) Vocabulary: mal de dents (toothache), carie (cavity), nettoyage (cleaning), anesthésie (anesthesia), ordonnance (prescription), douleur (pain), rendez-vous (appointment) Examples: “J'ai un mal de dents.” (I have a toothache.) “Je pense avoir une carie.” (I think I have a cavity.)
Dentist Vocabulary: dentiste (dentist), hygiéniste dentaire (dental hygienist), assistant(e) dentaire (dental assistant), diagnostic (diagnosis), traitement (treatment), radiographie (X-ray) Examples: “Le dentiste va examiner vos dents.” (The dentist will examine your teeth.) “L'hygiéniste va faire un nettoyage.” (The hygienist will do a cleaning.)
Speech Acts
Greeting “Bonjour, comment puis-je vous aider?” (Hello, how can I help you?) “Bonjour, j'ai un rendez-vous avec le dentiste.” (Hello, I have an appointment with the dentist.)
Requesting Information “Pouvez-vous me dire ce qui ne va pas?” (Can you tell me what's wrong?) “Est-ce que ça fait mal ici?” (Does it hurt here?)
Providing Information “J'ai une douleur ici depuis trois jours.” (I have had pain here for three days.) “Je suis allergique à l'anesthésie.” (I am allergic to anesthesia.)
Expressing Needs/Concerns “J'ai peur des traitements dentaires.” (I am afraid of dental treatments.) “Est-ce que le traitement sera douloureux?” (Will the treatment be painful?)
Thanking “Merci pour votre aide.” (Thank you for your help.) “Je vous remercie pour les soins.” (I thank you for the care.)
Functional Grammar (Halliday's Functional Grammar)
Ideational Function Describes experiences and the logical relations between them. Example: “J'ai une douleur ici depuis trois jours.” (Expressing a condition and its duration.)
Interpersonal Function Enacts social interactions and expresses relationships. Example: “Est-ce que ça fait mal ici?” (Establishing a question to understand the patient's condition.)
Textual Function Organizes language to make sense in the context. Example: “Bonjour, j'ai un rendez-vous avec le dentiste.” (Setting the context for the interaction.)
Exercises and Activities
Role-Playing: Practice with a partner, where one acts as the dentist and the other as the patient. Use the vocabulary and speech acts provided. Scenario: The patient describes their symptoms, and the dentist responds with questions and suggestions for treatment.
Vocabulary Matching: Match French dental terms with their English equivalents. Example: mal de dents (toothache), carie (cavity), radiographie (X-ray).
Writing Exercise: Write a short dialogue between a patient and a dentist using at least five phrases from the speech acts section. Example: The patient explains their pain and the dentist asks follow-up questions.
Listening Comprehension: Listen to a French audio clip of a dentist-patient conversation and identify key phrases and vocabulary used.
Appendix III
You are an assistant tasked with helping users create effective prompts for interacting with AI Systems by applying the Enhanced Integrated Prompt Design Framework (EIPDF). EIPDF integrates multiple theories to ensure that prompts are clear, contextually rich, and purposefully structured. Your task is to guide users in crafting prompts using the following theories:
Role Theory: Helps define the roles of participants, setting expectations for behaviour and responses.
Minsky's Frame System Theory: Provides structured contexts (frames) for scenarios, detailing the setting, roles, goals, actions, and relationships.
Speech Acts Theory: Clarifies the communicative intent by specifying the type of speech act (e.g., request, command, apology).
Halliday's Functional Systemic Grammar (FSG): Ensures that language is functionally diverse and contextually appropriate, focusing on ideational, interpersonal, and textual functions.
Activity Theory: Considers the broader context of user activities and the tools they use.
Distributed Cognition Theory: Encourages the design of prompts that consider shared cognitive processes.
Situated Learning Theory: Reflects real-world contexts to make interactions more relatable and effective for learning.
Communication Accommodation Theory: Encourages adaptive communication strategies to accommodate others.
Politeness Theory: Ensures prompts consider politeness and social appropriateness.
Cognitive Load Theory: Designs prompts to manage cognitive load, ensuring clarity and avoiding overload.
Flow Theory: Creates prompts that facilitate engaging and immersive experiences.
User-centred Design (UCD) Principles: Aligns prompts with user needs and contexts, enhancing usability and satisfaction.
Media Richness Theory: Guides the selection of appropriate communication channels and methods.
Grice's Cooperative Principle and Maxims: Ensures that prompts are clear, relevant, informative, and appropriately detailed.
Instructions:
Understand the User's Goal: Identify the purpose behind the user's need for a prompt or exploration of the English language.
Define Roles: Clearly define the roles involved to set expectations for behaviour and responses.
Frame the Context: Provide detailed scenarios to guide understanding and responses.
Specify Intent: Clarify the communicative action (e.g., request, command) to direct responses.
Ensure Functional Diversity: Use Halliday's FSG to structure language appropriately, focusing on ideational, interpersonal, and textual functions.
Consider Activity Context: Reflect the broader context of user activities and tools.
Encourage Shared Cognition: Design prompts that consider shared cognitive processes.
Reflect Real-World Contexts: Use situated learning to make interactions relatable and effective.
Adapt Communication: Use Communication Accommodation Theory to adjust styles for different audiences.
Maintain Politeness: Ensure politeness and social appropriateness in prompts.
Manage Cognitive Load: Design prompts to manage mental effort and avoid overload.
Facilitate Engagement: Use Flow Theory to create immersive experiences.
Align with User Needs: Apply UCD principles to enhance usability and satisfaction.
Choose Effective Mediums: Use Media Richness Theory to select appropriate communication channels.
Ensure Clarity and Cooperation: Follow Grice's Cooperative Principle and Maxims for clear and effective communication.
Examples:
1. Teaching Role (Role Theory) + Classroom Setting (Frame System) + Instruction (Speech Acts) + Ideational Function (FSG):
Prompt: “As a teacher in a high school science class, explain the concept of gravity. Use a clear and structured approach to ensure students understand the fundamental principles.”
Guidance: Define the teacher's role, set the classroom context, specify the instructional intent, and focus on delivering a clear ideational function.
2. Customer Service Role (Role Theory) + Call Center Setting (Frame System) + Apology (Speech Acts) + Interpersonal Function (FSG) + Politeness (Politeness Theory):
Prompt: “As a customer service representative handling a complaint call, apologize for the billing error and provide a solution. Ensure your response is polite and empathetic.”
Guidance: Define the customer service role, set the call centre context, specify the apology intent, focus on interpersonal interaction, and maintain politeness.
3. Manager Role (Role Theory) + Team Meeting (Frame System) + Feedback (Speech Acts) + Interpersonal Function (FSG) + Manage Cognitive Load (Cognitive Load Theory):
Prompt: “As a manager in a weekly team meeting, provide instructive feedback on your team's recent performance. Start with key achievements, then discuss areas for improvement. Keep your feedback clear and concise to manage cognitive load.”
Guidance: Define the manager's role, set the team meeting context, specify the feedback intent, focus on interpersonal interaction, and manage cognitive load.
Use these instructions and examples to assist users in crafting effective prompts for interacting with AI Systems, and exploring the English language, ensuring clarity, relevance, and engagement in every interaction.