iiinova — where your digital self keeps evolving

Kevin Feng @kjfeng.me · 27/05/2026

If you're attending ACM CAIS 2026 in San Jose, be sure to check out our new work led by Tanjal Shukla (@onaheater.bsky.social) on AI agents that can adapt their autonomy based on user interaction! Both Tanjal and @leijiew.bsky.social are attending, so pls stop by our poster and ask for a demo!

021

Kevin Feng @kjfeng.me · 01/08/2025

We also introduce **agent autonomy certificates** to govern agent behavior in single- and multi-agent settings. Certificates constrain the space of permitted actions and help ensure effective and safe collaborative behaviors in agentic systems.

120

Kevin Feng @kjfeng.me · 01/08/2025

Level 5: user as an observer. The agent autonomously operates over long time horizons and does not seek user involvement at all. The user passively monitors the agent via activity logs and has access to an emergency off switch. Example: Sakana AI's AI Scientist.

100

Kevin Feng @kjfeng.me · 01/08/2025

Level 4: user as an approver. The agent only requests user involvement when it needs approval for a high-risk action (e.g., writing to a database) or when it fails and needs user assistance. Example: most coding agents (Cursor, Devin, GH Copilot Agent, etc.).

100

Kevin Feng @kjfeng.me · 01/08/2025

Level 3: user as a consultant. The agent takes the lead in task planning and execution, but actively consults the user to elicit rich preferences and feedback. Unlike L1 & L2, the user can no longer directly control the agent's workflow. Example: deep research systems.

100

Kevin Feng @kjfeng.me · 01/08/2025

Level 2: user as a collaborator. The user and the agent collaboratively plan and execute tasks, handing off information to each other and leveraging shared environments and representations to create common ground. Example: Cocoa (arxiv.org/abs/2412.10999).

100

Kevin Feng @kjfeng.me · 01/08/2025

Level 1: user as an operator. The user is in charge of high-level planning to steer the agent. The agent acts when directed, providing on-demand assistance. Example: your average "copilot" that drafts your emails when you ask it to.

100

Kevin Feng @kjfeng.me · 01/08/2025

120

Kevin Feng @kjfeng.me · 01/08/2025

100

Kevin Feng @kjfeng.me · 01/08/2025

100

Kevin Feng @kjfeng.me · 01/08/2025

100

Kevin Feng @kjfeng.me · 01/08/2025

100

Kevin Feng @kjfeng.me · 01/08/2025

100

Kevin Feng @kjfeng.me · 01/08/2025

📢 New paper, published by @knightcolumbia.org We often talk about AI agents augmenting vs. automating work, but how exactly can different configurations of human-agent interaction look like? We introduce a 5-level framework for AI agent autonomy to unpack this. 🧵👇

Screenshot of a paper on a webpage, with a figure showing 5 levels of autonomy for AI agents and the tradeoff between user involvement and agent autonomy as the levels increase.

1204

Kevin Feng @kjfeng.me · 27/04/2025

Excited to kick off the first workshop on Sociotechnical AI Governance at #chi2025 (STAIG@CHI'25) with a morning poster session in a full house! Looking forward to more posters, discussions, and our keynote in the afternoon. Follow our schedule at chi-staig.github.io!

041

Kevin Feng @kjfeng.me · 25/04/2025

🤖LLMs are being integrated everywhere, but how do we ensure they're delivering meaningful user experiences? In our #chi2025 paper, we empower designers to think about this via 🎨designerly adaptation🎨 of LLMs and built a Figma widget to help! 📜 arxiv.org/abs/2401.09051 🧵👇

The first page of an academic paper with a figure depicting the use of Canvil, a Figma widget for designerly adaptation of LLMs, along with the names of the 6 authors.

1173

Kevin Feng @kjfeng.me · 21/04/2025

The First Workshop on Sociotechnical AI Governance at CHI 2025 (STAIG@CHI’25) is less than a week away! We’re super excited to hold a panel with some amazing speakers to discuss why a sociotechnical approach to AI governance is so important. Learn more + see our full schedule: chi-staig.github.io

poster advertising 5 panelists from a diverse set of institutions the first workshop on sociotechnical AI governance at CHI 2025

042

Kevin Feng @kjfeng.me · 16/01/2025

To make our paper more accessible to not only researchers, but also devs, journalists, etc. we have a FAQ in the Appendix! Be sure to check that out, and if you have more questions, please don’t hesitate to reach out.

A screenshot of the FAQ section in the appendix of the Cocoa paper.

100

Kevin Feng @kjfeng.me · 16/01/2025

🖥️Interactive UI Cocoa embeds AI agents into docs with a new design pattern—interactive plans. They’re “computational notebooks” for humans + agents, breaking tasks down into interactive steps. We designed Cocoa for scientific research, an important & challenging domain for AI.

A screenshot of the Cocoa user interface, with the following image description: An overview of the Cocoa user interface. An interactive plan affords human-AI co-planning and co-execution: a researcher and the AI agent can collaboratively edit the plan in the document and execute the plan steps, similar to executing code cells in a computational notebook. Steps can be assigned to the AI agent or the researcher. The researcher can freely edit the AI agent’s outputs in an interactive sidebar and add new information to help steer the agent with their feedback and expertise. In this example, the first three steps of a plan to summarize methods for human-feedback elicitation have already been executed, and the agent is requesting guidance from the user in the next step.

110

Kevin Feng @kjfeng.me · 16/01/2025

🤔Giving complex tasks to AI agents is easy—getting them to do exactly what you want isn’t. How can human-AI collaboration give us more reliable & steerable agents? 🍫Introducing Cocoa, our new interaction paradigm for balancing human & AI agency in complex human-AI workflows. 🧵

A screenshot of the first page of an academic figure with a list of authors and a high-level schematic figure describing Cocoa, an interactive system for co-planning and co-execution with AI agents.

2387

Kevin Feng @kjfeng.me · 23/12/2024

📢📢 Introducing the 1st workshop on Sociotechnical AI Governance at CHI’25 (STAIG@CHI'25)! Join us to build a community to tackle AI governance through a sociotechnical lens and drive actionable strategies. 🌐 Website: chi-staig.github.io 🗓️ Submit your work by: Feb 17, 2025

A poster advertising the first workshop on sociotechnical AI governance with a description of the workshop's core themes and faces of the organizers.

24517

Kevin Feng @kjfeng.me · 15/07/2023

Unbeatable pastries in Montreal's Little Italy @ Alati-Caserta

A shelf full of Italian pastries including different kinds of cannoli, bomboloni, and cakes.

010

Kevin Feng @kjfeng.me · 08/07/2023

Turns out custom feeds were around us this whole time!

020

Kevin Feng @kjfeng.me · 18/06/2023

Found authentic Guangdong-style chang fen (rice noodle rolls) in downtown Montreal! First time having these in 6 years, sooo good and brings back childhood memories 😋

A plate of thin, white rice noodle sheets wrapping together some meat and vegetables in a pool of dark sweet soy sauce.

010

Kevin Feng @kjfeng.me · 12/06/2023

Hellooo Montreal 🇨🇦

A large metallic ring situated between two buildings near St. Catherine's street in downtown Montreal

A street with light beige Victorian style townhouses. The townhouses have cartoon-style flowers painted into their first floor exterior walls with the words "fleurs d'asie du sud-est" in French and "flowers of South Asia" in English

A view of the Montreal skyline from a high vantage point, with some treetops in the foreground

A wide, busy pedestrian-only street during the day many people walking and some people sitting in front of the steps of a large concrete and glass building with colourful lettering and graphics.

010

Kevin Feng @kjfeng.me · 31/05/2023

But—there's a catch. Enabling provenance adds another dimension of credibility: how valid is the actual provenance info? We tested different states of provenance credibility and found that they had a significant impact on users' credibility ratings of *the content itself*. 4/n

3 different states of provenance: normal, incomplete, and invalid. The colours of the circular provenance indicators are black, orange, and red, respectively. Incomplete and invalid indicators also have a brief description of the state in the UI displaying the provenance chain.

010

Kevin Feng @kjfeng.me · 31/05/2023

We partnered with folks at C2PA and Adobe to create "provenance-enabled" prototypes of Twitter-like feeds, simulating UX outlined in the C2PA provenance standard. We then populated these prototypes with diverse content and deployed them to 595 ppl in the US and UK. 2/n

A social media post with an image, featuring a pop-up UI showing the edit history for that image.

010

Kevin Feng @kjfeng.me · 08/05/2023

Nice day of clamming! We dug around at low tide and ended up finding big patches of horse clams. Cleaning them was a different story tho...didn't expect to learn so much about clam anatomy in one day

Gray beach at low tide with plenty of seaweed on the ground against a blue sky with white clouds.

A group of clams sitting on gray sand and green kelp beside a large hole from which the clams were dug.

Top-down view of a pile of clams in a red bucket.

110

Kevin Feng @kjfeng.me · 04/05/2023

European muesli is incredible. Nothing like it

000

Kevin Feng @kjfeng.me · 23/04/2023

Iconic Swiss railway clock design is also used by Deutsche Bahn

000

Kevin Feng @kjfeng.me · 22/04/2023

First meal in Berlin: pork hock, schnitzel, beer 🇩🇪

010