Lab
Tools and interventions that support human judgement when people work with AI.
Introduction
Small by design. Tested in practice.
The Lab is where I build and test tools and interventions that support human judgement when people work with AI. Each starts with a question from real work, becomes a working prototype, and is tested to see what actually shifts.
Some lead to practical interventions. Others open up better questions.
BE Cue Detector
Making the Invisible Visible.
Conversational AI can sound knowledgeable, attentive and remarkably human. But the language that makes it believable isn't necessarily a reflection of its capabilities.
The BE Cue Detector uses AI to identify and highlight three kinds of textual cues: social, expertise and mirroring. By making these cues visible, it creates opportunities to notice, question and discuss how AI shapes our perceptions.
It grew out of my research. Then something unexpected happened: when people saw the cues highlighted in AI text, they started talking about what those cues might mean for their own conversations. That turned it into an intervention, an experiment in designing for human judgement.
The BE Cue Detector highlights social in yellow, expertise in green and outlines mirroring cues with a purple outline. In the tool, clicking a highlight explains why it was labelled.
BE Diagnostic
AI proposes. A person decides.
Automating the work was useful only if I could still evaluate what the system was doing.
The BE Diagnostic is an experimental research tool for identifying and examining textual cues that make conversational AI feel believable.
AI proposes labels for social and expertise cues and explains each one. A human reviewer accepts, rejects or revises them. I built it to support my research into the Believability Effect.
The tool exports labels, including each reviewer’s validation status and edits, to a CSV. That supports audit, helps refine the labelling rubric, and builds datasets for examining patterns across AI-generated responses.
Beyond research, the approach opens up possibilities for independent AI evaluation. Organizations could examine how vendor systems communicate, signal expertise and respond to users, rather than relying exclusively on vendor-provided assessments. It’s designed for use in advisory work.
The BE Diagnostic identifies and classifies textual cues in conversational AI responses, providing AI-generated explanations and allowing human reviewers to evaluate the results. Note: Colour coding is being aligned with the BE Cue Detector.
Where this started
Before conversational AI, I was asking the same question of other technologies: what happens when people meet them in the real world? These projects are where my approach comes from. Ask the people doing the work, start small, and learn from what happens.
Where’s Janneke? | 2018
Can an interdisciplinary team go from idea to working prototype in one week?
Three students, three disciplines and one challenge. Observe how people work, find a problem worth solving and build something useful.
They noticed colleagues were always looking for me. Their solution? An Amazon Alexa skill called "Where's Janneke?"
One week later, it worked.
And when they proposed tracking my whereabouts, I had one condition: they had to experience being tracked too.
Ask the people doing the work.
HestaHub | 2019
What if your home knew when to ask for help?
Developed as a concept for Collision 2019, HestaHub explored how existing smart-home technologies could help people living independently.
The idea was simple: keep personal data inside the home and notify a trusted person when something needed attention. No constant surveillance. No unnecessary sharing of personal information.
Built around a principle carried forward from earlier robotics experiments:
know when to ping a human.
Saffi & Elly | 2018-2019
What happens when social robots meet the real world?
Exploring what robot co-workers might mean for banking, we took Saffi and Elly out into the wild, from corporate events to TIFF and a Hermès pop-up.
We discovered that people had strong reactions to the robots, often wanting to interact or take photographs. Those encounters inspired new possibilities, including a social media partner and a talent scout to help connect young people with recruiters.
The experiments also exposed the limitations of the technology, particularly around data privacy, control and the unpredictable nature of human interaction.
The experience eventually led to a presentation at the National University of Singapore.
People respond to human-like machines before they think about them.
Big questions. Small experiments. Better ways forward.
I build purposeful interventions that put people at the centre. I use what already exists, create what doesn’t, and move quickly from an idea to something we can test.
Adopting AI and wondering where judgement belongs? Bring what you’re working on. 30-minute conversation, no pitch, no obligation.