Show Me the Data

The episode contains genuine technical substance around BBI/PPG signal processing, federated learning for biotech privacy, and the spike-recovery metric for emotional regulation, but these are interspersed with lengthy backstory, philosophical meandering about humanity and Silicon Valley, and platitudes about belonging. The density of actionable insight per minute is uneven.

“emotion isn't what we think it is. Societally, you know, we use kind of emotion and feeling interchangeably. But if I ask you how you're feeling and you say, I feel happy. M on an autonomic level, you've experienced about 50 emotions before you've even answered that question.”

“a perpetrator of violence goes into anger and it actually takes them about five hours to come back to a balanced state. And so what we can look out for is what's actually supporting that person to bring their spike recovery down to like 10, 15 minutes”

The federated learning-for-biotech framing and the analogy of standardising emotion the way humanity standardised time and language are genuinely fresh framings. However, much of the mental health system critique, Silicon Valley/transhumanism scepticism, and prevention-vs-crisis-spending argument are well-worn positions in this space.

“I kind of put it in the same vein as standardizing time, which has allowed us to do complex global trade, standardizing language, which has given us complex thought, philosophy, science, et cetera. But emotion, which is central to the human experience, we're all just kind of comfortable with. I don't really know what that is.”

“rather we open source it and say if you're interested in our ah, data, build on our infrastructure. Plug in Truth as a data layer into your university”

Gibson is a genuine practitioner - she founded and ran a nonprofit at 18, completed a two-year 300-community national tour, served as youngest-ever national mental health commissioner, and is now building an early-stage emotion biotech company with real partnerships (Garmin). She is not a career podcast guest. However, InTruth is clearly pre-scale and several claims are highly aspirational rather than demonstrated at volume.

“in that time there was 300 communities that um, my then team and I went to and tens of thousands of stories that we had heard in all different kinds of environments”

“the federal budget, which at the time was 27 billion. And when I stepped into that role, I said yes. And I served in that role for three terms.”

There are useful concrete details - $4/unit white-label PPG sensors, 3-to-7-day baselining period, the circumplex/valence-arousal model, Garmin as a named partner, 1-in-4 Australian families affected by family violence, and the 5-hour vs. 10-15-minute spike recovery framing. However, the 80% unconscious-emotion claim is asserted without citation, and several bold product claims lack supporting data or customer evidence.

“if we wanted to white label PPG sensors from China for about $4 a unit”

“It takes um, three to seven days and then the tech will continue to baseline you”

The host frequently answers his own questions, riffs alongside the guest rather than probing, and offers no pushback on large unverified claims (e.g., 80% of behaviour unconsciously driven by emotion, influencing geopolitical affairs via emotion data). The co-host's one technical question about hardware accuracy is the sharpest moment of the conversation.

“I like it, I like what you said. It's kind of like so human to human. It could help you have a argument with your spouse.”

“Yeah. And you could, by measuring it, could sort of give people warning ahead of. It's like measuring your heartbeat.”