TLDR: A message claiming to come from a human must genuinely come from a human. Bots impersonating their operators corrode trust in all communication.
Authentic human intent is the property of a message that was initiated, written, and sent by a human with genuine purpose. It is what separates “I’d love to meet” from a scheduling bot firing off a request under your name.
AI assistants are breaking this contract. People let bots compose and send emails in their own name — same signature, same “I”, same apparent decision. The recipient cannot tell. This is a problem.
Trust in asynchronous communication rests on one implicit assumption: if a message says it comes from Alice, Alice decided to send it. When bots violate that assumption silently, the rational response is to distrust everything. Every “I wanted to follow up” becomes suspect. Every expression of interest might be an automation. That epistemic corrosion is expensive and structural.
The fix is not to ban AI from communication. The fix is honest attribution. The sender of a message must reflect who actually initiated it.
I call this the bot twin pattern:
- Martin Monperrus sends an email → sender is
martin@monperrus.netwith an authentic human intent - Martin’s bot sends an email → sender is
monperrus-bot@monperrus.net, this is automation
The bot twin signals two things at once: this was not a direct human
decision, and the human operator is identifiable.
monperrus-bot tells the recipient who is responsible
without pretending to be human.
This applies beyond email. A post on social media, a comment on a pull request, a message in a chat thread — if a bot wrote it, the bot’s identity should appear. Not a human’s.
The rule is simple: a message claiming human intent must have authentic human intent. A bot twin is a commitment to a communication ecosystem where human signals still mean something.
See also: - AICID: identities for AI Scientists