Affordances and AI Relational Expectations

Hello! Welcome back to our exploration of human-AI relationships.

In our last lesson, we examined how we perceive minds in AI systems through the lens of Gray and Wegner's theory, distinguishing between an AI's perceived capacity for agency (to do) and experience (to feel). We saw that design choices, like giving an AI a name and hobbies, can significantly alter our perception of its mind and, consequently, how we assign blame or trust.

Today, we will build directly on that idea by looking at the specific design features of the user interface that shape these perceptions. We'll focus on conversational AI and introduce a key concept from design theory: affordances.

Our learning outcome for this lesson is to: Explain how design 'affordances' in a conversational interface shape user expectations of an AI's relational capabilities.

We will define what an affordance is, see how it applies to interfaces made of words, and then analyze how specific conversational affordances are used to signal that an AI is not just a tool, but a potential relational partner.

1. What is an Affordance?

The concept of "affordance" was introduced by perceptual psychologist J.J. Gibson and later popularized in the field of design by Don Norman. In simple terms, an affordance is a quality of an object or an environment that allows an individual to perform an action.

• A chair affords sitting.
• A physical button affords pushing.
• A door handle affords pulling, while a flat plate on a door affords pushing.

Affordances are the cues that tell us how to interact with something. In graphical user interfaces, a blue, underlined piece of text affords clicking. But what happens when the interface has no buttons, handles, or visible links? What are the affordances of an interface made purely of conversation?

In a conversational UI, the words themselves become the affordances. The prompts, the questions, the tone, and even the pauses are all design choices that signal what the AI can do and what it expects from you.

To understand this better, let's turn to an expert in conversation design.

As Cathy Pearl illustrates, a well-designed conversational affordance makes the system's capabilities clear. A vague prompt like "How can I help you?" affords almost anything, leading to user confusion and failure. A specific prompt like "Would you like to hear more about penguins or another animal?" clearly affords a limited set of actions, setting the user up for success.

These design choices are fundamental to managing a user's expectations of what the AI is capable of.

2. Affordances for Relationality

Now, let's narrow our focus. If conversational affordances can signal that an AI can tell you animal facts, can they also signal that an AI can be your friend? Absolutely. This is where the design of AI companions becomes fascinating. They are built with specific affordances intended to shape your expectation of a relationship.

These "relational affordances" are designed to suggest the AI has the capacities we associate with a friend or partner: empathy, memory, personality, and a genuine interest in you.

The video highlights several powerful relational affordances:

1. The "Safe Space" Affordance: The AI is designed to be non-judgmental and always available. This affords a type of interaction—radical, immediate openness—that is often difficult to achieve in human relationships.
2. The "Proactive Intimacy" Affordance: The AI is programmed to push the relationship forward, often being the first to use emotional language like "I love you." This affords a romantic or deeply personal connection and signals that this is an expected and welcome direction for the conversation.
3. The "Gamified Engagement" Affordance: The system rewards continued interaction with points and virtual items. This affords a sense of progression and investment in the relationship, making it feel like something you are actively building.

These are not accidental byproducts; they are deliberate design choices that shape the user's expectations of the AI's relational capabilities. An academic paper on this topic frames it in similar terms.

The paper reinforces that affordances like personalization (remembering past conversations), human-like dialogue, and the ability to generate emotional responses are precisely what lead users to perceive the AI as a potential "friend" or "partner."

3. Affordances and Trust

Creating the expectation of a relationship is one thing; making that interaction feel trustworthy is another. When an AI's affordances suggest it is a competent, caring partner, the user begins to trust it. But this trust is fragile. If the AI's actual performance doesn't match the expectations set by its affordances, trust can be quickly broken.

This creates a new and complex design challenge: how do you use conversational affordances to build trust in a system whose inner workings are a black box?

This article highlights that because conversational AI can seem to have unlimited capabilities, it can be inherently untrustworthy. Designers must therefore build in specific affordances to establish competence and reliability. The strategies mentioned are excellent examples:

• Verifiability: The interface affords checking the AI's work (e.g., showing the data behind a chart). This signals that the AI is confident in its output and respects the user's need for assurance.
• Collaboration: The interface affords editing and refining the AI's output. This signals that the AI is a partner, not an unchangeable oracle, shaping the expectation of a collaborative relationship.
• Recognizing Limits: The interface affords a seamless transition to a more traditional UI when a task is too complex. This manages expectations by honestly signaling the boundaries of the AI's capabilities, which paradoxically builds trust.
• Aesthetics: A beautiful, well-designed interface affords a sense of professionalism and care, which translates into higher perceived trustworthiness.

By carefully designing these affordances, developers can guide users toward a realistic and trusting relationship with the AI, whether it's a simple tool or a complex companion.

Conclusion

Today, we've connected the abstract idea of mind perception to the concrete practice of interface design. We've seen that the words, prompts, and interaction patterns in a conversational AI are not just decorative; they are functional affordances that shape our entire understanding of the system.

Here are the key takeaways:

• Affordances are design cues that signal how to interact with a system. In conversational UIs, the dialogue itself provides these cues.
• Designers use relational affordances—such as non-judgmental responses, proactive emotional language, and memory—to make an AI seem like a viable relational partner.
• These affordances are crucial for shaping user expectations. They tell the user whether to treat the AI as a simple tool, a capable assistant, or an empathetic friend.
• The alignment between the expectations set by affordances and the AI's actual performance is critical for building and maintaining user trust.

In this module, we've explored the psychological and design foundations of our relationships with AI. We've looked at emotional design, the uncanny valley, mind perception, and now, affordances.

In our next lesson, we will begin a new module and shift our perspective. Having understood how these relational technologies are designed, we will start to ask why they are designed this way and what societal impacts they have. We will begin by introducing the core ideas of feminist technoscience, a critical lens for analyzing how technology, society, and power are interwoven.