Running a workshop

This guide is for anyone who wants to run these lessons with a group of people, in a village, a school, a university or an office. It describes how GainForest runs them in the field, most recently at Parque das Tribos and Arte & Escola na Floresta near Manaus in 2026. Change anything that does not fit your place.

Before you go

Be invited, and agree on the purpose. A workshop that arrives to validate a decision someone else already made is a consultation in disguise. If you are not from the community, your job is to host the questions. The answers belong to the community. Early in her statement on Indigenous data, Audrey Tang notes that a formal position on Indigenous policy "should be reviewed, shaped and led by the appropriate indigenous institutions." Bring the same humility into the room.

Work with local champions. At Parque das Tribos, Samuel and Eduardo set up the equipment with us and ran it after we left. At Arte & Escola na Floresta they came along with Marina Mura and translated and guided the sessions. A tool introduced by someone who already uses it lands differently from one introduced by visitors.

Pay people for their time. At the January 2024 workshop at Parque das Tribos, each of the more than thirty participants was paid for taking part. Budget for food as well.

Check the small barriers. New Telegram accounts in Brazil needed a R$7 SMS payment, and some people could not sign up. Bring spare phones with the apps already installed and logged in. Check power, cables, adapters and the connection the day before.

Decide what you will record. Ask each person before you photograph, record or transcribe them, and accept a no. What the workshop should produce is the group's agreements, such as the three-column sheet, the council charter or the assistant's personality card.

Fit the schedule to people's lives. We held the Manaus BioBlitz after school hours so that young people could join.

Ask for funding. GainForest funds one community workshop or BioBlitz each month through the GainForest Workshops grant. Apply with your plan and pick Data is Soil as the topic.

A two-day plan

This plan follows the 2026 workshops in Manaus, with the lessons that fit each part. A three-day BioBlitz followed, where groups went out on their own with Taina and came back to compare what they had found.

Day one: in the room

• Setup with the local champions. Starlink, Raspberry Pi, phones. Explain to everyone that what is on the device belongs to the community.
• Opening round. Each person says their name, their community, one word connected to the forest and the story behind it. At Arte & Escola we asked for a movement instead of a word, and everyone stood up.
• Lesson 01. Train a model on your own faces, then look at the GBIF map.
• First encounter with Taina. Hand out the phones and let people explore before you explain anything.
• Lesson 03. What stays public, what stays close, what is left alone.
• Lesson 08. Who will sit on the data council, and how they are chosen.

Day two: outside

• Walk with Taina. Photograph and identify what you find.
• Lesson 11. Put out two or three AudioMoth recorders where the group decides.
• Lesson 13. Draw the place first, then fly the drone and compare.
• Back inside. Show everything the group collected on a screen and let people find their own records. At Parque das Tribos the room filled with "that one is mine."
• Feedback wall. Two columns on a large sheet: what you liked and what you did not. Write them down word for word.
• Lesson 10. Redesign the assistant's personality together.
• Hand over the kit and set the date of the next meeting.

Shorter formats

• Two hours: lessons 01 and 03.
• Half a day: lessons 01, 03 and 05, or 01, 03 and 10 if you have phones with Taina.
• A technical team: lessons 02, 04, 05 and 07. Run the refusal lab with real engineers in the room.
• Funders and partners: lessons 01, 06, 09 and 12.

In the room

• Let people answer first. Ask why the model failed before you explain unbalanced data. People usually get there on their own.
• Leave disagreements on the wall. If the group cannot place an item in one column, that is useful information. Do not force a vote.
• Let others lead. At Arte & Escola the children chose the tree for the recorder and swam across the river to tie it up. That was not in the plan and it was the best part of the trip.
• Follow up on what hurts. When people talk about knowledge taken without consent, or about a snakebite, slow down. These are not data questions.

Did anything change?

Counting participants and hours shows reach. To see whether people learned something, our community lead uses three questions, asked before and after a session, with a board and markers:

• Know. What do you understand about this?
• Can. What do you feel able to do?
• Do. What have you actually done or practised?

In Semporna, Sabah, teenagers who had survived fires answered "not sure" when asked what they could do about disaster preparedness. Two hours later they had drawn three community risk maps. Often people already know a great deal and need a way to name it.

After you leave

• Leave the tools. The kit stays with the local champions, with written instructions in their language.
• Come back. The communities near Manaus told us that researchers had often visited once and never returned. Coming back with changes people asked for is what built trust.
• Fix what they told you. After Parque das Tribos we fixed a bug that put gallery photos in the wrong place, made uploads survive a dropped connection, and started working on editing observations after upload.
• Write back. Tell us what worked, what the room refused and what you changed: team@gainforest.net.

Mistakes we have made

• We assumed a free app meant free access, until the Telegram SMS fee.
• We expected voice messages to work. Speech recognition did not understand local dialects.
• Uploads that worked on good WiFi failed as soon as the signal dropped.
• Early on we used forms and categories like "plants" and "animals" that did not match how people describe the forest.