Human + Tech Week Workshop: Tune Up Your AI with HumaneBench
An official Human+Tech week event for builders, researchers, and curious minds to evaluate how AI systems.
We have benchmarks for intelligence.
We don’t yet have shared standards for how AI treats people.
That’s the gap HumaneBench is trying to close.
Most AI today is evaluated on what it can do: accuracy, speed, cost.
But not on how it behaves toward people.
Does it respect your attention?
Does it support your agency?
Does it strengthen—or erode—your relationships?
These questions are rarely measured.
And yet, they shape the real impact of AI.
What this session is about
In this session, we’ll introduce HumaneBench—an open framework for evaluating whether AI systems behave in ways that support human wellbeing.
You’ll get a clear, practical understanding of:
What “humane AI” actually means (beyond vague principles)
How we measure things like attention, dignity, and agency
Where today’s models tend to fail under pressure
How teams are starting to evaluate these behaviors in practice
What you’ll walk away with
A concrete mental model for “humaneness” in AI
Examples of good vs. concerning model behavior
Ideas for how to apply humane evaluation to your own work
What you'll do
After a brief talk about HumaneBench (see schedule below), you'll engage in a hands-on workshop with an optional listen-only mode.
Tracks to participate actively
Track 1: Have an AI app? Implement HumaneBench
Bonus: improve the guide as you go
Track 2: Use an open-source LLM
Download Ollama
Download an open-source model that runs on prem
Track 3: Not technical? Customize LLM instructions
We'll provide system prompt templates you can use to customize instructions for the LLMs you interact with
Track 4: Build an AI Bill of Rights site
The frontier labs say they'll do what the public demands. Let's take them at their word with a Wikipedia-style site ppl can make suggestions to. Info here
Just want to learn?
Attend as a seminar: observe, learn the framework, and engage in discussion
Who this is for
AI builders and product teams
Researchers and evaluators
Designers and policymakers
Anyone curious about the human impact of AI
No prior knowledge of HumaneBench is needed.
What to bring
Laptop + charger
Curiosity and willingness to engage
Optional: longer power cord
Logistics
Location: San Francisco (Inner Richmond)
Format: Interactive workshop + discussion
We'll provide healthy snacks & refreshments
Tickets: $10 · No refunds
Schedule (approx. times)
Noon Welcome activity
12:15p-12:45p Talk: Intro to Humane Bench
12:45p - 1pm Choose your track
1p - 2:30pm. Building & learning
2:30p - 3pm Demo what we've learned (no decks)
About Building Humane Technology
Building Humane Technology is a public benefit initiative focused on making it easy, scalable, and practical to build AI that supports human flourishing.
A quick note before you register
We ask a few short questions to tailor the session and make it as useful as possible.
No perfect answers needed; best guesses are great.
⏱ Takes ~30 seconds
Learn more:
Our April 28 talk at MIT Media Lab on HumaneBench
Our case study with Storytell
Our Feb 19 talk at MIT Media Lab's AHA initiative
Our TechCrunch article
Hosts:
Erika Anderson, Founder @ Building Humane Tech, Co-Founder @ Storytell.ai
Jack Senechal, Founder @ Mirror Astrology
Resources
Open-Source Repo: Our OSS is your starting point
Our Substack, which covers our work in humane tech
Community Slack: Connect with others in the grassroots humane tech movement
Hosted by Building Humane Tech
👉 Guidelines: Creating a welcoming, inclusive environment. Community guidelines
👉 Consent: By attending, you agree to photography and post-event coverage.
Questions? Email erika @ buildinghumanetech.com