Prompting Dialogue | Styles and User Behavior
Jonathan "Kyle" Hobson
UX Researcher | Human Factors | ASU Master's Graduate | Curating AI-Infused Human Experiences
Classification of Prompting Styles
In the domain of human-AI communication, the interplay between user behavior and the effectiveness of AI interaction is heavily influenced by the styles of prompting employed. Prompting styles, as they are broadly defined, shape the nature of the outputs generated by AI, reflecting a spectrum of user engagement levels, strategic foresight, and interaction goals. This section introduces the concept of "Prompting Styles," setting the stage for a detailed exploration of their classification and impact on AI communication.
Prompting styles can be broadly categorized based on the approach users take in interacting with AI systems:
The diversity in prompting styles not only illustrates the varied strategies employed by users but also highlights the nuanced understanding required to optimize interactions with AI systems. Each style embodies a unique set of behaviors, preferences, and interaction patterns that contribute to the broader taxonomy of user engagement with artificial intelligence.
By classifying prompting styles, we aim to correlate these methods with user effectiveness and satisfaction, offering insights that could guide the development of more intuitive AI interfaces and interaction protocols. This initial exploration sets the groundwork for deeper analysis into how different prompting styles affect the dynamics of human-AI communication, potentially transforming our approach to AI system design and user education.
Figure 3.1.1: Prompting Preferences Across Experience
Prompting Styles Across Experience Levels in AI Interaction
Note. Represents survey results from "Exploring the Landscape of Prompting" survey and Usability Tests conducted by Jonathan Kyle Hobson. Shows data distribution of AI prompting styles across varying experience levels, highlighting trends in stylistic preferences among novice to expert professionals. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Novice vs. Expert
Among the spectrum of AI prompters, a pragmatic approach is markedly favored, with 25% of novice users and a significant portion of expert professionals aligning with this style. Interestingly, a detailed methodology is prevalently adopted by both expert professionals and experienced users, while a high number of novices lean towards minimalist prompting. These trends suggest that as proficiency grows, so does the inclination towards more intricate and explorative prompting strategies, potentially reflecting a deeper understanding and nuanced application of AI interaction techniques.
Figure 3.1.2: Prompting Preferences Across Experience
Correlation of AI Prompting Styles with Model Preferences
Note. Represents survey results from "Exploring the Landscape of Prompting" survey and Usability Tests conducted by Jonathan Kyle Hobson. Shows data categorizing AI users by their preferred prompting styles and correlates these with the AI models they predominantly use, from minimalist to detailed prompters. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Shows distinctive affinity for minimalist prompters towards Pi, a conversational AI model. Conversely, users of ChatGPT, Claude, Llama, and Gemini exhibit a tendency towards a detailed prompting style. This correlation may indicate that certain AI models, with their unique capabilities, attract and are better suited to particular prompting techniques, with ChatGPT and similar platforms appealing to users who engage in more complex, detail-oriented prompting.
Prompting Styles to Personas
As we delve deeper into the classification of prompting styles within the realm of human-AI communication, it becomes clear that these styles can be aggregated into broader categories, encapsulating complex user profiles that we will refer to as personas. These personas provide a nuanced view of user interactions with AI, blending various prompting characteristics into relatable archetypes that reflect common usage patterns and technological preferences.
The move from individual prompting styles to the concept of personas allows us to present a more dynamic and comprehensive understanding of user engagement with AI systems. By examining these personas, we can better tailor AI systems to meet the diverse needs of users, enhancing both the effectiveness and the user experience of AI interactions.
Detailed Exploration of Personas
The Efficiency-Oriented persona combines elements of the Minimalist, Utility-Driven, and Pragmatic styles. This persona is characterized by a preference for AI tools that offer straightforward solutions and integrate seamlessly into existing workflows, optimizing productivity without excessive interaction. Users fitting this persona typically engage with AI for task-specific purposes, seeking tools that enhance efficiency with minimal disruption to their daily routines. Their interaction with AI is driven by the need for quick, effective results, as seen in their frequent use of direct and concise prompts. The effectiveness of this approach is illustrated in Figure 3.2.1, which showcases typical interactions and the preferred AI tools that align with this persona's goals.
Figure 3.2.1: Efficiency-Oriented
Efficiency-Oriented Persona: Synthesis of Minimalist and Pragmatic Prompting Styles
Note. This persona encapsulates the efficiency-oriented approach, merging minimalist, utility-driven, and pragmatic prompting styles into a coherent profile. Created by researcher Jonathan Kyle Hobson using Figma, it is based on comprehensive survey data and observational research aimed at optimizing AI interactions for streamlined utility.
In contrast, the Engaged Interactive persona reflects a combination of Iterative, Pragmatic, and Contextual prompting styles. Users in this category utilize AI as an integral part of their creative or problem-solving processes, often engaging in a continuous feedback loop to refine outcomes. This persona values AI tools that can provide dynamic, context-sensitive support, enhancing both creativity and productivity. Their use of AI is characterized by a balanced approach to depth and efficiency in information processing, which is effectively depicted in Figure 3.2.2. This figure highlights how these users strategically employ AI to achieve a blend of innovative and practical results.
Figure 3.2.2: Engaged Interactive
Engaged Interactive Persona: Pragmatic and Contextual Prompting in Harmony
Note. This figure presents the engaged interactive persona, a fusion of pragmatic and contextual prompting styles, designed to foster a more dynamic interaction with AI systems. The persona, crafted in Figma by Jonathan Kyle Hobson, reflects a user type dedicated to richer, more nuanced AI engagements, informed by empirical research findings.
The Creative Explorer persona merges the Detailed, Explorative, and Enthusiast styles into a profile that seeks to expand the boundaries of what AI can achieve. This persona is deeply involved in using AI for complex problem-solving and innovation, often in academic or research settings. They favor open-ended and highly detailed prompts that challenge both the AI and their own analytical skills. The interaction patterns of this persona, as seen in Figure 3.2.3, emphasize their preference for AI tools that support high-level cognitive processes and creative exploration.
Figure 3.2.3: Creative Explorer
Creative Explorer Persona: Detail-Oriented and Exploratory Prompting Fusion
Note. Depicted here is the creative explorer persona, integrating detailed and explorative prompting styles, with an added zest for enthusiast-driven interactions. Formulated by Jonathan Kyle Hopson in Figma, this persona is for users who delve deeply into the possibilities offered by AI, pushing the boundaries of creativity and discovery.
So far we have been looking at the AI professional prompter, the enthusiast, or the individual who may call themselves the casual prompter and while these prompters will “bridge the gap between” the “end users and the large language model" (Amazon Web Services, 2024) they will not be the average end user. The persona Casual Adopter is that average end user. One who primarily aligns with the Minimalist style but may occasionally incorporate other styles when necessary. This average user will use AI when it is integrated into their existing applications, typically without knowing they are doing so in straightforward tasks that require minimal input and yield immediate results. This persona "avoids trial and error” yet “still wants to receive coherent, accurate, and relevant responses from AI tools" (Amazon Web Services, 2024). The simplicity and accessibility of AI tools are paramount for these users, who prefer technologies that are easy to use and integrate into their existing routines without additional complexities. Figure 3.2.4 illustrates this persona's typical AI usage scenarios, emphasizing their preference for simplicity and efficiency. This persona shows us that when UX Designers or Prompt Architects are building these systems we build for the user not the prompter.
"AI must be designed with the user in mind—every interaction, every response should feel like a step towards a simpler, more efficient life" (Hulten, 2021)
Figure 3.2.4: Casual Adopter
Casual Adopter Persona: The Minimalist Approach to Everyday AI Use
Note. The casual adopter persona embodies a minimalist prompting style, suitable for users who may engage with AI inadvertently in various applications. Jonathan Kyle Hopson has developed this persona in Figma to characterize users who prefer simplicity and effortlessness in AI-powered tools, often without conscious interaction with the underlying technology.
By analyzing these personas, as detailed in the accompanying figures, we gain valuable insights into the diverse ways users interact with AI, from those seeking efficiency without complexity to those exploring the creative and cognitive possibilities of advanced AI interactions. This understanding is crucial for developing AI systems that are not only technologically advanced but also deeply attuned to the varied needs and preferences of users across different contexts.
Building Intelligent Experiences: User-Centric AI Design
As we continue to explore the landscape of AI prompting and user behavior, it is crucial to consider Hulten (2018) principles that guide the development of intelligent experiences. These principles are essential for creating AI systems that not only deliver accurate and valuable outputs but also prioritize user satisfaction and engagement. "Each successful interaction with an AI system should add noticeable value to the user's experience, outweighing the risks of potential errors" (Hulten, 2021). When designing AI-powered tools and applications, UX designers and prompt architects must carefully consider the following key aspects:
"Forcefulness of interaction should be tailored to the delicacy of human attention, not to overpower but to effectively guide" (Hulten, 2021).
By carefully balancing these Hulten (2018) principles, designers can create AI experiences that seamlessly integrate into users' lives, providing value and convenience without causing frustration or disruption. This user-centric approach is particularly important when catering to the needs of the Casual Adopter persona, who seeks simplicity and efficiency in their AI interactions.
As we navigate the complex landscape of AI prompting and user behavior, keeping these principles at the forefront of our design decisions will be essential for creating AI systems that truly resonate with users and deliver meaningful, intelligent experiences.
Interweaving Conversational Design
As the field of AI prompting evolves, the principles and practices of conversational design continue to play a pivotal role in shaping the way humans interact with artificial intelligence. Conversational design, the art of creating natural and engaging dialogues between users and systems, is integral to the development of effective AI prompting strategies. By leveraging the insights and techniques from this discipline, prompters can create more user-friendly, inclusive, and impactful interactions with AI systems across various waves of AI.
At the heart of conversational design lies the concept of natural conversation processing and understanding. As Greg Bennett, Director of Conversation Design at Salesforce, notes, "Conversation design is fundamentally the practice of designing interaction flows strategizing forms of language like specifically what is said and how it said" (Bennett & amUX, 2022). This involves designing AI systems that can comprehend and respond to human language in a way that feels intuitive and organic. By incorporating elements such as discourse markers, turn-taking, and conversational repair, prompters can create AI interactions that mimic the flow and nuances of human conversation. As Greg Bennett emphasizes, "Those little tiny words, while often they get dismissed as filler words actually have a very big role in conversation" (Bennett & amUX, 2022), highlighting the importance of these subtle elements in crafting natural AI interactions.
"Conversation as a sort of interaction mechanism, if you will, it's pretty easy to mess up it's something that I think we often sort of take for granted" (Bennett & amUX, 2022). This underscores the complexity of human communication and the challenges in replicating it through AI. As Cathy Pearl, Head of Conversation Design Outreach at Google, points out, "when we're talking to people so much more is going on than those individual words. We've got eyegaze, we've got body language, we've got pauses, we've got intonation" (Google et al., 2019). Even brief pauses can convey significant contextual information, as Pearl illustrates: "If the pause between our turns is longer than that, that is information. I say, 'Can I get a ride tomorrow morning?' and more than a second goes by, something's up... This was only one second of time and yet so much information is communicated in that time" (Google et al., 2019).
"Language is also ambiguous, so let's say you say to me, 'Do you want coffee?' and I say, 'Coffee will keep me awake.' Show of hands how many people here think I just said no I do not want a cup of coffee... well, it depends right if it's this morning I'm probably gonna say yes, if it's right before bed this probably means no." (Google et al., 2019)
To navigate these complexities, prompters must develop interaction models or architectures that account for a variety of utterances, intents, and entities, guiding AI systems to provide accurate and contextually relevant responses. This is particularly crucial in action-driven conversations, where the goal is to lead users towards specific tasks or outcomes through natural, goal-oriented dialogue. As Pearl explains, "Effective conversation design… relies on adhering to the cooperative principle” and Maxim of Relevance, “assessing… these systems maintain contextually appropriate interactions" (Google et al., 2019).
Inclusivity is another key consideration in conversational design for AI prompting. "It's very important to make sure that you... understand these different versions of English, because... you have whole populations" (Bennett & amUX, 2022). By embracing linguistic diversity and avoiding linguistic bias, prompters can create AI experiences that are accessible and welcoming to a wide range of users.
The principles of conversational design extend beyond text-based interactions, encompassing voice user interfaces (VUIs) and multimodal systems. As AI evolves through its various waves and systems, prompters must adapt their strategies to accommodate emerging technologies like subvocalization and silent speech. By crafting conversations that seamlessly integrate multiple modes of communication, prompters can create immersive and engaging AI experiences.
By taking a close look at conversational design we don’t reinvent the wheel in AI prompting. We can see the importance of investing effort in crafting clear, concise, and contextually relevant prompts. By doing so, prompters can ensure that AI systems can more easily understand and fulfill intents. This not only enhances the user experience but also enables AI systems to provide more accurate and valuable responses.
In conclusion, conversational design remains a critical component of AI prompting, regardless of the wave or system of AI being employed. By leveraging the principles and techniques of this discipline, prompters can create AI interactions that are natural, engaging, and truly transformative. As the field of AI prompting continues to evolve, staying informed about the latest developments and best practices in conversational design will be essential for crafting AI experiences that are not only technologically advanced but also deeply human-centered.
“Is the glass half-empty, or is the glass half-full? I think about this question often. I know it's supposed to be a question of pessimism or optimism, but I think it's really a question that requires more context. You see, whether the glass is empty or full depends heavily on the state of the glass in the first place. If you have an empty cup, and I fill it half-way, you would likely say that your cup is half-full. However, if you have a full cup, and I empty half of it onto the ground, you would likely say that your cup is half-empty. My point is that we can't compare glasses without comparing context. You and I might feel entirely differently about the same amount of water in our cup, and both of our feelings are justified. I know it's supposed to be about perspective, but don't forget that perspective is deeply affected by our experience, so dig deeper, ask questions, and don't assume that everyone's glass is the same” (Hanson, 2023).