The Tweeting Badger

Read the campaign story

A zoo’s daily life contains stories, but turning those stories into a steady social presence takes staff time. Johannesburg Zoo’s 2013 Tweeting Badger project gave its honey badger BG a social persona linked to activity in the enclosure. Sensors triggered prewritten messages related to where the animal moved. The badger was not composing language; a human-designed system translated activity into character-led updates.

ITWeb’s contemporary report describes motion sensors, a base station and a server selecting messages. The campaign archive credits Hellocomputer and Draftfcb, with BinarySpace involved in the technology. The system’s connection to real movement distinguished it from an entirely fictional mascot account. At the same time, the language remained authored, so the persona should not be mistaken for a literal account of the animal’s intentions.

The mechanism was a small stream of physical events becoming recurring content. Instead of manufacturing a new idea for every post, the team created rules that turned relevant activity into an ongoing voice. This can make a destination feel alive between visits. The creative task moves from writing one post to defining a character, event categories and limits on repetition.

A small organisation can apply the principle to non-sensitive operational data: a workshop might show a completed repair count or a garden might share a verified bloom update. Explain what is automated and keep human oversight. Never create activity merely for the content, especially where animals or people are involved. The system should document normal work rather than distort it for engagement.

Follower counts in contemporary reporting do not establish extra visits or conservation impact. The opened sources also do not provide an independent welfare assessment. The practical lesson is to connect content to a real, renewable source of events while remaining transparent about the translation. Automation can reduce repetition in production, but editorial judgement and responsibility cannot be delegated to a sensor.

Use real event inputs

Connect recurring content to something that genuinely happens.

Disclose the translation

Explain that automated messages are human-authored interpretations.

Keep human oversight

Monitor repetition, errors and unintended effects.

A small test for another brief

Publish transparent updates from a harmless recurring operational event.

Accuracy, repetition rate and useful audience responses.

Stop if content production alters animal care or exposes private information.

• Choose an event that occurs without staging.
• Write a small, clearly automated message set.
• Review output and add a human stop control.