Exploring the Effect of Anthropomorphic Conversational Interfaces on UX and Adoption in Crypto Wallets
Blockchain adoption lags behind other emerging technologies because wallets are difficult to understand and don’t align with users’ existing mental models. When faced with complex or unfamiliar systems, users often rely on what they already know, which can make blockchain interactions feel confusing and untrustworthy. Conversational interfaces and subtle anthropomorphic cues provide a way to simplify these experiences, enhance clarity, and foster the trust necessary for broader adoption.
Lack of feedback during loading states
Poor loading states where transactions appear to be ‘pending’ for long periods without context or visual feedback worsen the user experience, making it difficult for users to understand whether the system is functioning correctly.
Complex technical barrier
The limited design of loading screens in many crypto wallets often forces users to navigate complex, unfamiliar technical language and rely on external blockchain explorers to verify transaction status.
Erroneous mental model
Crypto wallet developers do not consider existing user mental models and therefore do not design interfaces that align with users' expectations. Research suggests that users rely on traditional financial mental models to fill knowledge gaps when sending or receiving crypto assets, often comparing their experiences to banking applications.
Poorly designed blockchain wallets create confusion and uncertainty, which directly undermines trust. Prior research shows that user trust in blockchain technology is strongly tied to the quality of the user experience when interacting with wallets and related tools. When interfaces are unintuitive, users report lower confidence and higher frustration, factors that discourage adoption. As a result, poor UX has become a primary barrier to behavioral intention to adopt blockchain-related tools and services.
Emergent Needs
Transparent feedback on blockchain processes
When users send a transaction, they need to understand what is happening in real time. Wallets should provide clear, step-by-step feedback that reduces uncertainty, such as displaying pending confirmations, estimated completion times, or the current status of a transaction within the blockchain process.
Simplified learning of crypto concepts
Most users lack technical knowledge of blockchain, which makes wallets intimidating. Interfaces should gradually introduce key terminology and processes in plain language, giving users the confidence to act without feeling overwhelmed.
Wallets built around user mental models
People approach new technologies with expectations shaped by familiar apps and services. Crypto wallets should align with traditional mental models, offering familiar design patterns, user flows, and predictable interactions that reduce uncertainty.
Conversational and human-like guidance
Complex tasks become easier when guided by a conversational interface. Adding chat-based support or subtle human-like cues can make the experience feel more approachable, improve clarity, and build trust through relatable interactions.
While MetaMask's transactional experience is designed to enable users to send or receive cryptocurrency transactions, two additional interfaces will be created: one that integrates a conversational interface, and another with a conversational interface that incorporates anthropomorphic elements.
RQ1. How do users’ trust, attitude, user experience, and intention to adopt a crypto wallet differ across a conventional interface, conversational interface, and anthropomorphic conversational interface after completing a cryptocurrency transaction?
RQ2. How do participants’ perceived trust, comprehensibility, and security change across a pending state with progress and contextual feedback versus minimal feedback, and do these changes differ depending on the type of crypto wallet interface?
Traditional interface
This user flow was designed to mimic MetaMask's traditional desktop wallet interface for sending assets. The goal of this control group was to establish a baseline experience that reflects what users typically encounter in the current market. This version provides only the minimal feedback cues common to existing wallets.
Curent state
User flow
Conversational & anthropomorphic wallet
This user flow was designed to integrate a conversational agent into the wallet experience, simulating a dialogue with the user. Instead of relying solely on static forms, the agent provides step-by-step guidance throughout the transaction process.
Incorporated changes
User flow
Traditional MetaMask wallet
Conversational & anthropomorphic wallet
Design feedback loops that build trust
Participants felt more confident, trusting, and knowledgeable after receiving contextual feedback with progress indicators. Designers should prioritize guiding users through the process and providing useful contextual information that can enhance participants’ understanding of blockchain.
Include visual feedback showing transaction stages.
Explain why delays happen and what’s occurring behind the scenes.
Share educational cues in real time (e.g., “Waiting for validator approval”).
Don’t rely on conversational interfaces alone
Although conversational and anthropomorphic UIs were explored, they did not significantly improve trust, usability, or adoption intent on their own. Many users were unsure whether the agent's feedback was honest. If the agent’s behavior lacks responsiveness or interactivity, users may see it as inauthentic or gimmicky.
Conversational interfaces must clearly respond to user actions in real time.
Static or non-responsive agents can create mistrust and confusion.
Lessons learned
Verbal and visual feedback cues can significantly improve how users perceive blockchain technology and their perception of time.
Users consistently rely on mental models when transacting with crypto wallets.
When dealing with emerging technology, less experienced users benefit from process guidance, which may lead to higher levels of trust.
Conversational agents are often linked with LLMs and AI, and therefore can trigger uncertainty.
