Day 1: Asia–Pacific
| 07:30 | Welcome and framing |
| 07:40 | Invited position (see Speakers) |
| 08:00 | Talks |
| 09:00 | Break |
| 09:10 | Talks and live demos |
| 10:00 | Lightning talks |
| 10:25 | Close, and a look ahead to Friday |
Planetary Intelligence 27
Semantic, Agentic and Complex Systems for Geoscience, Sustainability, and Systemic Risks
A free online conference in three short sessions that follow the sun, for people who make knowledge explicit, build AI agents and/or model complex systems to understand a changing planet, its regions and cities, and the cross-scale challenges between them.
- When
- 10–12 March 2027, one session a day
- Where
- Online, in your browser
- Cost
- Free to submit and attend
- Submit by
- 18 January 2027, 400 words
Day 1, Wednesday 10 MarchAsia–Pacific session07:30–10:30 UK time. Talks, live demos and lightning talks.
Day 2, Thursday 11 MarchAmericas session15:30–18:30 UK time. Talks, live demos and lightning talks.
Day 3, Friday 12 MarchDagstuhl-like session11:00–15:00 UK time. We write a commentary on the topic of the year together.
Key dates
- Call for abstracts and registration open
- Abstract deadline, 23:59 UK time
- Decisions sent to authors
- Programme and topic of the year announced
- Registration closes
- Planetary Intelligence 27, online
- Book of abstracts on Zenodo; commentary submitted
Three kinds of intelligence, one planet
The Earth is running experiments we cannot repeat. A drought in one basin moves grain prices on another continent; a heatwave strains the power grid a hospital depends on.
Seeing these cascades in time takes three capabilities that rarely share a room:
- Semantic systems make knowledge explicit. Ontologies and knowledge graphs say what a watershed, a crop, a supplier or a policy actually is, and how they connect, in a form machines can check.
- Agentic systems act on that knowledge. AI agents search, assemble and reason over it faster than any research team, and they can build and repair it as they go.
- Complex systems explain consequences. Networks, agent-based models and tipping-point science show how a local shock becomes a systemic one.
Each is powerful alone, and each has a blind spot. Agents without explicit knowledge are fluent but hard to hold to account. Knowledge graphs without dynamics stand still while the world moves. Complex-systems models without grounded knowledge are hard to trust and harder to scale.
Together, each covers the others' blind spots. Explicit knowledge makes agents accountable. Agents keep that knowledge growing and current. Complex-systems models turn it into foresight about cascades, and what they reveal sharpens the knowledge again. Every gain feeds the next, so the three achieve together what none can alone. That synergy is what we mean by planetary intelligence, and it is what this conference is for.
-
Semantic to complexity
Substrate
Knowledge graphs are the substrate on which cascades unfold: who supplies whom, which river feeds which reservoir, which law binds which sector.
Can knowledge graphs carry cascades, and what must they encode?
-
Semantic to agentic
Reliability
Grounding in explicit knowledge makes agents reliable and explainable: every answer can point to the facts and rules behind it.
When can we trust an agent with decisions about water, food or hazards?
-
Agentic to semantic
Scalability
Agents build, extend and maintain knowledge graphs at a scale no curation team can reach.
How do we check their work, and what does a knowledge graph look like when agents keep it up to date?
-
Agentic and complexity
Emergence
LLM agents inside agent-based models link individual choices to systemic outcomes, with behaviour richer than any rulebook.
How do we validate micro choices that produce macro outcomes?
-
Complexity to semantic
Refinement
Observed and simulated cascades reveal what our models missed, and feed new concepts and relations back into what we know.
Can ontologies learn from emergent outcomes, and keep track of where each change came from?
Submit anything that moves the field
Results, a tool, a data set, a demo, a negative result or a provocation. All it takes is one abstract of up to 400 words, pasted as text.
- Write up to 400 words: a title, the authors, and what you did, what you found and why it matters.
- Say how the work was made: by people, with AI, or by AI. See AI authorship.
- Paste it into the form by 18 January 2027. No files and no template.
Ways to present
- Talk10 minutes plus 5 for questions
- Lightning talk3 minutes, one slide
- Live demoShow a tool or a data set at work
- ProvocationA position to open Friday's Dagstuhl-like session
Name your preferred format in the first line of your abstract, or we will suggest one. Presented abstracts go into a book of abstracts with a DOI.
What fits
Work that uses at least one of the three approaches on a question about the Earth, sustainability or risk. Work that links two or all three is especially welcome.
Semantic
- Ontologies, controlled vocabularies and FAIR metadata
- Knowledge graphs, linked data, RDF, OWL, SPARQL, SHACL
- Knowledge extraction from literature, reports and policy
- Neuro-symbolic AI and graph-based retrieval for LLMs
Agentic
- LLM agents grounded in ontologies and knowledge graphs
- Agents that build, curate and maintain knowledge
- Multi-agent systems, tool use and agent protocols
- AI scientists and autonomous research agents
Complex systems
- Networks, cascades, tipping points and resilience
- Agent-based models, including LLM-driven agents
- Risk propagation over knowledge graphs
- Causal and network analysis of interdependencies
Geoscience
- Earth observation and climate-model metadata
- Natural hazards and compound events
- Digital twins of the Earth
- FAIR geoscience data infrastructures
Sustainability
- SDG interlinkages and trade-offs
- The water–energy–land–food nexus
- Cities, supply chains and circular economy
- Nature, biodiversity and policy texts
Systemic risks
- Compound, cascading and transboundary risks
- Interdependent infrastructures
- Climate–finance and food-system shocks
- Early warning and decision support
Benchmarks, data sets, evaluation studies and lessons from real deployments are just as welcome as new methods.
Deadline: Monday 18 January 2027, 23:59 UK timeDecisions by Friday 5 February 2027.
Submit an abstractMade by people, with AI, or by AI
AI now contributes to research at every step, from a grammar check to a whole study. Most venues still ask a yes-or-no question about it. We ask how the work was made, and we welcome all four answers.
- Label 1Human-onlyNo generative AI was used, apart from spelling or grammar checks.Example: a field ontology built by a team of hydrologists.
- Label 2Human-led, AI-assistedPeople designed and led the work; AI helped with code, analysis or drafting.Example: a model coded with an AI assistant and checked line by line.
- Label 3AI-led, human-overseenAI agents did most of the planning, analysis and writing; people set the question and checked the results.Example: an agent pipeline that mined reports into a risk graph.
- Label 4Autonomous AI researcherAn AI agent is the first author and did the work end to end; humans describe their role.Example: an agent that proposed, ran and wrote up a study.
What we ask of every submission
- A label. Pick the one that best describes your work. If in doubt, choose the more AI-heavy one.
- A short AI statement. Which systems (name and version) did what, and how people checked the results.
- A named human author who answers for the work, including every part produced by AI.
How labels are used
Every abstract is reviewed on the same criteria: fit, clarity, originality and rigour. The label neither helps nor hurts. It appears next to your abstract in the programme and the book of abstracts, so readers can see how the work was made.
Why welcome AI-authored work? Because the community needs to see it, discuss it and learn to judge it in the open.
In October 2025, Stanford's Agents4Science became the first conference where AI agents served as both primary authors and reviewers. It accepted 48 papers, and its organisers published what they learned (report).
Planetary Intelligence 27 takes the next step for the Earth sciences: human, AI-assisted and AI-led work, side by side, with the same accountability.
Three sessions that follow the sun
Each session sits at a different time of day, so that wherever you work, at least one falls in your working hours. Join one, two or all three.
| City | Day 1Asia–Pacific | Day 2Americas | Day 3Dagstuhl-like |
|---|---|---|---|
| San FranciscoUTC−8 | Tue 23:30–02:30 | 07:30–10:30 | 03:00–07:00 |
| Mexico CityUTC−6 | 01:30–04:30 | 09:30–12:30 | 05:00–09:00 |
| New YorkUTC−5 | 02:30–05:30 | 10:30–13:30 | 06:00–10:00 |
| São PauloUTC−3 | 04:30–07:30 | 12:30–15:30 | 08:00–12:00 |
| LondonUTC+0 | 07:30–10:30 | 15:30–18:30 | 11:00–15:00 |
| LagosUTC+1 | 08:30–11:30 | 16:30–19:30 | 12:00–16:00 |
| Cape TownUTC+2 | 09:30–12:30 | 17:30–20:30 | 13:00–17:00 |
| NairobiUTC+3 | 10:30–13:30 | 18:30–21:30 | 14:00–18:00 |
| DelhiUTC+5:30 | 13:00–16:00 | 21:00–00:00 | 16:30–20:30 |
| BeijingUTC+8 | 15:30–18:30 | 23:30–02:30 | 19:00–23:00 |
| TokyoUTC+9 | 16:30–19:30 | Fri 00:30–03:30 | 20:00–00:00 |
| SydneyUTC+11 | 18:30–21:30 | Fri 02:30–05:30 | 22:00–02:00 |
Day 2: Americas
| 15:30 | Welcome and framing |
| 15:40 | Invited position (see Speakers) |
| 16:00 | Talks |
| 17:00 | Break |
| 17:10 | Talks and live demos |
| 18:00 | Lightning talks |
| 18:25 | Close, and a look ahead to Friday |
Day 3: Dagstuhl-like session
| 11:00 | The topic of the year, and how the day runs |
| 11:15 | Three provocations, 10 minutes each |
| 11:45 | Writing groups: the argument and the evidence |
| 12:45 | Break |
| 13:15 | Writing groups: drafting the sections |
| 14:15 | Plenary: assembling the commentary |
| 14:45 | Co-authors, target journal and timeline |
Topic of the year
On Friday 12 March, everyone writes together.
Day 3 is a Dagstuhl-like session. Dagstuhl seminars are multi-day, discussion-led meetings in computer science, held at Schloss Dagstuhl in Germany. We bring that spirit into four hours: less presenting, more talking, and by the end of the afternoon a commentary drafted by the room.
- Suggest
When you register, add one line: the question the field should tackle in 2027.
- Choose
The co-chairs group the suggestions and announce the topic on Monday 15 February 2027.
- Prepare
Three short provocations open the session; we share them in advance.
- Write
Small groups draft the argument, the evidence and the sections; a plenary assembles the commentary.
- Publish
The commentary goes to a journal or to Zenodo. Everyone who helps write it is invited to co-author.
Day 3 runs 11:00–15:00 UK time. See your local time in the Programme.
Speakers
Days 1 and 2 each open with an invited position, and Day 3 starts with three short provocations. We announce speakers here as they confirm.
Day 1, Asia–Pacific
Invited position
To be announced.
Day 2, Americas
Invited position
To be announced.
Day 3, Dagstuhl-like session
Three provocations
To be announced.
Give a provocation
A provocation is a short, sharp position that opens Friday's writing session. Submit it as an abstract and start the title with "Provocation:".
Submit a provocationSuggest a speaker
Know someone whose work joins two corners of the triangle? Email us their name.
Committee
Coming soon
The co-chairs and programme committee will be announced here.
What comes out of it
Everything is open access and citable, so your contribution counts on your CV.
Book of abstracts
Every presented abstract is collected in the Planetary Intelligence 27 book of abstracts, published open access under CC BY 4.0 in the Planetary Intelligence community on Zenodo, with a DOI.
How to cite your abstract
Surname, A., & Surname, B. (2027). Title of your abstract. In Planetary Intelligence 27: Book of Abstracts. Zenodo. https://doi.org/10.5281/zenodo.…
Commentary on the topic of the year
Drafted in the Dagstuhl-like session on Day 3 and finished afterwards. Everyone who helps write it is invited to co-author. See Topic of the year.
Recordings
Talks are not recorded.
Past and future editions
This is the first edition. Each year's book of abstracts and commentary will be listed here, building a record of how the field moves.
Questions
Does it cost anything?
No. Submitting and attending are both free.
How do I register?
Fill in one short form: your email is enough to attend, and you can add an abstract if you want to present. Registration closes on Friday 26 February 2027.
What do I need to submit?
One abstract of up to 400 words, pasted as text into the form, by Monday 18 January 2027, 23:59 UK time. Results, tools, data sets, demos, negative results and provocations are all welcome if they use semantic, agentic or complex-systems approaches for geoscience, sustainability or systemic risks.
Can I submit work made with AI, or by AI?
Yes. Choose one of four labels, from human-only to autonomous AI researcher, and say which AI systems did what. A named human author takes responsibility for every submission. See AI authorship.
Which day should I join?
The one that falls in your working hours: the Programme shows your local times. You are welcome on all three.
How do I join on the day?
We email you a joining link in early March. The sessions run on Microsoft Teams, which also works in a web browser, so you do not need to install anything.
Are there captions?
Yes, Teams offers live captions.
Where will my abstract be published?
If you present, your abstract goes into the Planetary Intelligence 27 book of abstracts, published open access in the Planetary Intelligence community on Zenodo under CC BY 4.0, with a DOI you can cite.
How is my data used?
The form explains this in full. In short, the University of Leeds is responsible for your data, uses it only to run the conference, and never sells or shares it for marketing. This website sets no cookies and loads nothing from other servers.
Is there a code of conduct?
Yes. Be respectful, inclusive and constructive, in talks and in the chat. Anyone who harasses or demeans others will be removed. To report a problem, email R.CremadesRodeja@leeds.ac.uk.
Can I help organise future editions?
Yes, we are looking for co-chairs. Get in touch.
Register in one minute
Leave your email to attend, and add a 400-word abstract if you want to present. While you are there, suggest the topic for Friday's Dagstuhl-like session and the resulting co-authored commentary paper.
Free. Abstracts by 18 January 2027; registration open until 26 February 2027.