Exploring new interfaces for human-AI interaction

A growing list of 00 human-centered AI concepts that help designers align human needs with AI capabilities.

Thanks! Look out for your first update on human-AI design next Wednesday.
Oops! Something went wrong while submitting the form.
No items found.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Model management
Aug 2024
Automatic model switching

Automatic model switching in AI can boost efficiency by selecting the most appropriate model for each query, ensuring a balance between quick and accurate responses.

User feedback
Aug 2024
Quick or detailed feedback

Empower users to make decisions and give feedback quickly or engage more deeply when needed in natural language.

Controls
Aug 2024
Realtime voice interfaces

Voice interfaces should dynamically adapt to user interruptions, seamlessly incorporating them into the conversation ensuring a fluid and responsive dialogue.

Grounding sources
Jul 2024
Cross-tab assistance

A smart browser assistant that understands the context of your open tabs to offer relevant suggestions and actions, enhancing productivity through transparency and control.

A screenshot showing a web browser with multiple tabs open. The left side displays a sidebar with tabs for Gmail, a Medium article titled "Alignment through interaction," a YouTube video titled "Courage | The Art of Facing Fear," and a Slack channel named "AI Insights." The right side of the screen shows a Medium article featuring an iceberg image.
Templates
Jul 2024
Generate quick replies

An intelligent assistant that analyzes emails to identify questions and feedback requests, providing pre-generated response options and converting them into complete and contextually appropriate replies.

A screenshot of a message template interface, starting with a greeting to "Mike." The message confirms attendance for an event next Friday and requests Mike to bring either a dessert or some appetizers. Below the message, quick response options are available, allowing the user to confirm attendance and select the item they can bring: "Yes," "No," "Appetizer," or "Dessert."
Token builder
Jul 2024
Prompting with vibes

Input design concepts in small bits and see the cumulative output in real-time. Explore different combinations and immediately visualize the results, making the creative process interactive and flexible.

An image of a low poly, minimalistic game environment is shown with tags identifying it as a Western saloon. The scene is depicted as empty, featuring a landscape with trees and rock formations. The visual concept map includes tags such as "Game environment," "Western saloon," "Low poly," "Minimalism," and "Empty."
Templates
Jun 2024
Generative UI

Generative AI can provide custom types of input beyond just text, like generated UI elements, to enhance user interaction.

A user interface for selecting a movie, with a text conversation at the top. The user asks for help choosing a movie, and the assistant provides options based on preferences such as movie length and genres including action, comedy, romance, and others.
Summary
Jun 2024
Live chat echo

AI can enhance live chat streams by analyzing real-time data, identifying trends, and driving interactive elements like voting to boost audience engagement.

A screenshot from a live streaming platform showing a colorful game scene with multiple characters on an obstacle course. The chat sidebar is filled with enthusiastic messages reacting to a play, with one highlighted comment reading, "LMAO that was insane, crazy 10/10 play!"
Inline action
Jun 2024
Interactive writing partners

AI collaboration agents can act as writing partners that assist people by enhancing their content through transparent, easily understandable suggestions, while respecting the original input.

A screenshot of a document titled "The Impact of Artificial Intelligence on Modern Healthcare." It shows a paragraph about AI in healthcare, with a sidebar conversation offering editing suggestions, such as adding a definition of AI and expanding on predictive analytics.
Synthesis
Jun 2024
Navigate the space

Ordering content along different interpretable dimensions, like style or similarity, makes it navigable on x and y axes facilitating exploration and discovery of relationships between the data.

A collection of abstract, black ink sketches of human faces and figures. The central image is a close-up of a man's face with intense, angular lines. Surrounding sketches depict various male and female figures in expressive, fragmented styles.
Filters
Jun 2024
Semantic highlights

Embedding models can rank data based on semantic meaning, evaluating each individual segment on a spectrum to show its relevance throughout the artifact.

Highlighted text from a digital document titled 'Philosophical Perspectives on Love.' The highlights include various philosophical views on love, such as Plato's concept of love as a motivation to seek beauty and truth, and Aristotle's categorization of love into philia (friendship), eros (romantic love), and agape (unconditional love). The document suggests that love is a spectrum of relationships and emotions that influence human behavior and societal structures.
Show the plan
Jun 2024
Describe processes

AI actions often take time to complete. To improve user experience, use descriptions of what is happening combined with basic animations that represent different types of actions.

A user interface screen displaying four dark gray action buttons with icons and partial text. The actions are: 'Recognizing... usage patterns in your history,' 'Recommending... new articles based on your day,' 'Ranking... feedback from Peter by relevance,' and 'Generating... a response to the latest mail from Lisa.' The background shows faint, grayed-out icons and text for additional actions like 'Optimizing,' 'Visualizing,' 'Translating,' 'Testing,' and 'Summarizing.
Token builder
Jun 2024
Exploring language spatially

Use a spatial dimension to explore and manipulate language. By pulling text around on a map, you can play with different features in a playful and meaningful way.

 This image showcases a visual representation of different sentence structures and their characteristics. It displays four versions of a sentence about a boat trip in Venice, highlighting their differences in length, positivity, concreteness, and other attributes. The sentences are: "We had a beautiful boat trip in Venice" (1.27x Concrete). "Together we had such an unforgettable boat trip experience in Venice Italy" (1.41x Longer). "We had a pretty nice boat ride in Venice" (1.12x Negative). "Generating..." (1.35x Shorter). At the bottom, there are controls labeled Longer, Shorter, Negative, Positive, Abstract, and Concrete, indicating possible modifications to the sentences.
Synthesis
Jun 2024
Comparing embedding shapes

Comprehend and compare large documents by visualizing embeddings and their scores, enabling a clear and concise understanding of vast data sources in a single, intuitive visualization.

This image displays a radar chart comparing different characteristics of text embeddings. The categories measured are technical, fun, abstract, open, and personal. The highlighted text, "Collaborative building for a common culture," shows its position on the radar chart, which covers various extents in these categories. The chart is color-coded, and several colored areas overlap, indicating the different degrees to which these characteristics are present in the text. The interface includes a search bar with the mentioned categories as filters.
Variations
Jun 2024
Evolving outputs

Generating multiple outputs and iteratively using selected ones as new inputs helps people uncover ideas and solutions, even without clear direction.

The image shows a stylized flowchart with pink rectangles connected by lines, each containing a white heart icon, set against a gradient pink background. The chart appears to be moving in a staggered, diagonal direction from top left to bottom right.
References
Jun 2024
Language as a tangible material

Textual information often misses intuitive cues for understanding relationships between ideas. AI can clarify these connections, making complex information easier to grasp quickly.

An image displaying a text excerpt from an article with highlighted sections in beige. The highlighted text reads, "The new policy aims to reduce greenhouse gas emissions and combat climate change" and "However, the policy also includes provisions that encourage the continued use of fossil fuels, which are a major contributor to greenhouse gas emissions." To the right, a sidebar titled "The surprising twist #2" explains that despite the policy's aim to reduce emissions, it surprisingly includes provisions supporting the use of fossil fuels.
Synthesis
Jun 2024
Semantic ranking

Embedding models can rank content along virtually any dimension. This capability provides significant value by enabling users to explore and analyze the embeddings to create a spectrum of any features.

The image depicts a user interface element that appears to be a feedback sorting feature. The interface is designed with a pink background and a white overlay containing the sorting options and feedback content.
User feedback
Jun 2024
Evaluate predictions

When an observation is added to the context from an implicit action and a prediction is made, users should be able to easily evaluate and dismiss it.

An image showing a section of a user interface with a dark-themed background and purple dialogue boxes containing different suggestions or prompts. The background features a game or media title "Thunderstrike: Echoes of the Storm" with lightning and storm imagery. The suggestions include "Done browsing?", "Invite whole team?", "Organize tabs?", "Getting sleepy?", and "Not in the mood for Rock?". Below these prompts, a media control interface is visible with icons for infinity loop, skip back, play/pause, and favorite.
Data ownership
Jun 2024
Manage your data context

You should control how products and services (not) access your data through a manageable profile. This allows you to create a relevant context across many platforms while maintaining control.

An image showing a permissions request screen with a white background and green gradient header. The request is from OpenAI, asking to access a user's "datavault." The request includes checkboxes for "Personal information" (5 items), "Interests" (12 items), "Assistant preferences" (6 items), and "Contacts" (87 items). Each section has a dropdown arrow for more details, and there is a prominent "Allow" button at the top right.
Token builder
May 2024
Spatial Prompting

Spatial prompting integrates spatial relationships into prompts, offering a novel approach to manipulate concepts. This dynamic approach can lead to more intuitive and creative outcomes.

A web of tags around a text that says 'generating...' A curser hovers over one of the tags as if it is dragging it closer to or further from the middle.
Tone of voice
May 2024
Translate in X terms

Provide relatable and engaging translations for people with varying levels of expertise, experience and ways of thinking.

A screenshot of a translation tool with a dropdown menu showing "Chess game terms" and a purple "Translate" button below. The background text is blurred, focusing on the translation interface.
Synthesis
May 2024
Substantiated findings

LLM’s are great at organizing narratives and findings. It's helpful to see the sources that support these conclusions, making it easier to understand the analysis and where it comes from.

The image displays a grid layout with six cards, each containing an interview excerpt under the theme "Accessing music offline is difficult." Each card highlights a user's experience with offline music access issues: Interview with Emily: Emily finds it annoying when she can't access downloaded music due to a lack of signal or Wi-Fi. Interview with Rob: Rob is frustrated by being repeatedly asked to log in to access offline playlists during flights. Interview with Megan: Megan struggles to access offline content without an internet connection, defeating the purpose of offline mode. Interview with Bec: Bec relies on offline mode for commuting, but often finds songs not downloading properly or disappearing from her library. Interview with Thomas: Thomas appreciates offline mode but wishes for a feature to queue up songs in advance to avoid manual downloads in low connectivity areas. Interview with James: James often faces issues with the app not recognizing downloaded songs, leaving him without music in areas with spotty internet.
Follow up
May 2024
Follow-up on an answer

Letting people select text to ask follow-up questions provides immediate, context-specific information, enhancing AI interaction and exploration.

A screenshot highlighting an interface element labeled "Ask follow-up" appearing in a tooltip above text discussing the AGPTek 1080p Full HD Media Player.
Preview
May 2024
Realtime image generation

Realtime generation allows people to manipulate content instantly, giving them more agency in using generative AI as a tool for exploration.

This image displays a 3D glossy robot toy with a round blue body and a white face, featuring rosy cheeks and a smile. The toy has two black antennae on top of its head. The interface shows different views of the toy on the right side, and the caption below the main image reads, "3D tiny glossy robot toy."
Memory
May 2024
Proactive agents

Proactive agents can autonomously initiate conversations and actions based on previous interactions and context providing timely and relevant assistance.

This image shows a chat interface with three messages on an orange background. The first message, in a blue bubble, says "Great advice!" at 22:04. The second message, in a dark gray bubble, reads "Of course, That’s what I’m here for. See ya!" also at 22:04. There is a notification for "1 unread message" followed by the third message, in another dark gray bubble, asking "Did everything go well yesterday?" at 11:11.
Inline action
May 2024
Generative context menus

Based on your selection and situation, context menus can help you discover actions and access them quickly.

A screenshot of a text excerpt titled "Philosophical Perspectives on Love," with a highlighted section discussing Plato's views on love as a motivation for individuals, referencing his "Symposium." To the right, a context menu with options like Copy, Print, Read aloud, and other suggestions is visible. The background is pink.
Variations
May 2024
Incentivize giving feedback

Presenting multiple outputs helps users explore and identify their preferences and provides valuable insights into their choices, even enabling user feedback for model improvement.

This image features four abstract visual representations, possibly generated by Midjourney, showcasing "diffusion AI grain." The visuals include a central spiral structure, a flowing wave-like formation, a complex geometric mesh, and a field of wheat. The highlighted selection indicates "V2" with a cursor pointing to it.
Formula
May 2024
Prompts with nested data

Referencing nested data from your database in the form of tags can simplify the creation of elaborate prompt formulas.

An interface with an input field saying: ' Choose out of the [tags] for each... There is a field with multiple options to select a reference to nested data.
Suggestions
May 2024
Conversation starters

Starting with a blank canvas can be intimidating, but providing prompt starters can help individuals overcome this initial hurdle and jumpstart their creativity.

The image shows a user interface with various buttons labeled "conversation starters." On the right side, there's a panel labeled "GIF · Animation Studio" by witlist.design. This panel provides information about a tool or app, including a rating of 3.0, the DALL·E category, and mentions that it has been used in over 5,000 conversations. Some example conversation starters are "Spooky dance old cartoon," "Extreme gorilla drumming," "Meme funny scream," and "Robot dance performance." On the left side, there are search prompts like "Is AI gamification new?" and "AI prototyping tools?"
Filters
May 2024
Classifying topics

AI excels at classifying vast amounts of content, presenting an opportunity for new, more fluid filter interfaces tailored to the content.

This image shows a user interface with tabs for different categories at the top, such as "Topics," "Interface," "Accessibility," "Customization," and "Feedback." The "Interface" tab is currently highlighted in purple. Below these tabs are three sections representing different comments or discussions, each containing a circular user avatar icon and placeholder text that represents user comments or feedback. The background has a light purple hue.
Preview
Apr 2024
Realtime prompt feedback

Guide users to understand what makes a good prompt will help them learn how to craft prompts that result in better outputs.

A interface with feedback on a prompt being weak, average or great with recommendations to add context, role, format, tone, examples and constraints.
Summary
Apr 2024
Pinch to summarize

In Arc, a playful pinch interaction lets you quickly distill any webpage into a brief summary, capturing the essence of the content in moments.

The image depicts a two-panel interface design on a pink to purple gradient. The left panel resembles a web browser displaying a prompt for email updates. The right panel shows a summary of the website's content about human-AI interaction, key UX insights, and transparency. A curved arrow between the panels suggests content summarization with the phrase "Summarizing for you..." below.
References
Apr 2024
Visual grounding

Using the source input as ground truth will help trust the system and makes it easy to interpret its process and what might have gone wrong.

A screenshot of a pricing table or spreadsheet. The main focus is a highlighted cell displaying the amount "€ 35,00". This same amount is also shown in a separate box or label within the image. The surrounding cells contain various other monetary amounts ranging from around €27 to €165. Based on the layout and formatting, this seems to be a financial or pricing-related document.