Reading the honest signals of every living thing — with AI.
A plant has no face and no voice. But voltage runs through its leaves, and it changes when you touch it, when the light shifts, when you come near. This seminar teaches you to read that language — and the honest signals of humans, animals, and crowds — with machine learning you build yourself.
Mars is escape — the idea that if we ruin this planet we start over somewhere else. Learning to read the other living things that share it is the opposite: integration, learning to live as part of Earth's system rather than its exploiter.
This isn't science fiction. Cheap sensors, fast models, and the same machine-learning trick that cracked image recognition now let us reach signals that were always there and never readable — a face's micro-expressions, a cat's call above our hearing, the voltage in a leaf. AI here is a new sense organ.
You'll work with real research data and, in the best case, contribute to publications. No prior deep-learning experience is required. The path is deliberately scaffolded: everyone reaches a working result, and there's a steeper track for those who want it.
The course text — Hidden Signals: From Paracelsus to Plant Sensors — carries the whole arc, from a rebel physician who burned his textbooks to a mathematician who read the future of a crowd.
Emotion and stress read from expression, tone, and the rhythm of give-and-take.
Cat calls, dog and horse emotion — a language never meant as language.
Leaf voltage that answers to touch, sound, light — and human presence.
Collective mood from millions of messages, laid against events and markets.
Read nature directly, not the old books. Every body wears its inner state as a signature.
Knowledge shared freely across borders — the seed of open science, Wikipedia, open source.
You can't predict one person. A crowd of millions bends into a smooth, readable curve.
Perfect understanding across every species. Could a machine be that translator?
You'll form Collaborative Innovation Networks — small, self-organizing teams built on 20+ years of research into how real innovation happens, from Linux to Wikipedia to the Transformer behind ChatGPT.
The finding, in one line from our study of sixteen medical online communities: it is rotating leaders who build the swarm. Teams grew when leadership moved from person to person, when there were clear connectors, and when people wrote simply.
And we practice what we measure. We use AI to analyze your team's own communication — who answers whom, who connects, whether one voice pulls everything toward itself — and mirror it back to you.
That mirror alone changes teams. Groups shown how they collaborate become more open and attentive; in a two-year field study, their customer satisfaction measurably rose. Measurement here isn't surveillance — it's an invitation to self-knowledge.
Rotating leadership, fast responsiveness, balanced contribution, honest sentiment, strong connectors, shared attention — measured for a whole team, not just one person.
Draw the living network of a group from its communication patterns and watch it change over time — the same map, in motion.
The state where a team clicks. You'll learn to recognize it, measure it, and steer your own team toward it across the term.
Hands-on, scaffolded, guided step by step. You finish with something that runs.
Deeper methods, real research questions, publication potential.
Turn a leaf's signal into a spectrogram and let a model read touch, light, and nearby human mood.
Map who connects whom in a real open-source project — centrality, rotating leaders, the hidden hubs.
Emotion from a dog's or horse's posture; compare the model's guess with your own read.
Does the collective mood of a swarm forecast tomorrow's market or vote? Test it.
Diarize and analyze a real team meeting; mirror its network back and measure the change.
The boldest activity — and the sharpest ethical edge. Build one, then argue its limits.
// These are starting points. Teams propose and shape their own projects.
// Status-meeting dates and the final session are set at kick-off — update the two amber items then.
To join, contact the local administrator at your university before the start date. Work through Hidden Signals: From Paracelsus to Plant Sensors to get familiar with the concepts, tools, and research questions, then complete the five steps below and present your results at the block course. You can join any project on the Projects page.
→ Complete all five steps before the block course (14.10). Steps 1–5 take roughly 60–90 minutes in total.
Read Hidden Signals, then claim one chapter in the shared sign-up sheet. Teams are two students (a few chapters need three, flagged in the sheet); first come, first served. Your team presents and leads discussion on that chapter at the block course — you're the resident experts on it.
WhatsApp — the Symbiont Analyzer is now built into Beecome. Export a group chat you're active in (Menu → More → Export chat, without media) and load it in the app. You get a mix of the five collaboration archetypes — Bee, Ant, Butterfly, Capybara, Leech — plus the word patterns behind it. Note yours, and come ready to say whether it fits.
On-phone · interactive — no upload. The app plays a short video while your front camera reads your micro-expressions, and returns your dominant emotion and a predicted Big Five profile. Note your result and one thing that surprised you.
On-phone · interactive — no questionnaire. Swipe through a branching story; the model infers your Big Five from your choices. Compare it with your Perceptiface result — where do the two models agree, where do they diverge?
One row per student. Fill in each tool's output — archetype and percentage, the labels and personality profiles, your dominant emotion — and leave blank whatever a tool didn't show.
Identify your chapter's single most important claim, its strongest evidence, and where it meets the seminar's themes — honest signals and swarm creativity. Empirical chapters need at least one slide on the method or data.
Prepare exactly three discussion questions:
"Here is what the tool said about us. Here is what the chapter predicts it should say. Here is where they disagree — and that disagreement is our discussion question."
MIT Research Affiliate · Honorary Professor, U Cologne · 25+ years of COINs research.
peter.gloor@uni-koeln.de Profile →Professor of Information Systems & Social Networks, University of Bamberg.
Profile →Doctoral researcher in Information Systems, University of Cologne.
Profile →The English edition, chapter by chapter — the text this seminar runs on.
The German original the seminar is built on — the complete manuscript as a PDF.
Lab manuals and step-by-step build guides for the activities.
The self-analysis suite you'll run in the pre-work: Beecome — which now includes the Symbiont Analyzer — plus Perceptiface and Happimeter. Read your own communication, personality, and mood from real data.
The open plant-sensor research program behind the book: an ECG-style sensor reads a plant's bioelectric signals, and AI correlates them with human presence and emotion. You'll build the sensor and read its first voltage curve yourself.
Optional companion reading for the Master track.