Stop externalizing the cost of your AI use to me
20 days ago
- The author refuses to engage with AI-generated content that is sent without editing, as it externalizes the cost of reading and verifying to the recipient.
- Externalizing costs occurs when one party's actions impose burdens on others, such as oversized cars endangering pedestrians or companies privatizing profits while socializing losses.
- With AI, producing polished text became cheap, but the effort to read and provide feedback remains high, breaking the heuristic that effort signals value.
- Sending unedited AI output with disclaimers like 'just skim it' transfers labor without consent, undermining trust and feedback systems.
- This pattern harms open-source maintainers (overwhelmed by AI-generated issues), educators (exploited resubmissions), and peer review (half-baked papers using reviewers' labor).
- The author now becomes suspicious of AI-generated text, causing collateral damage to genuine efforts that are indistinguishable from lazy AI shortcuts.
- The proposed solution is a retreat to small trust networks, where norms like apologizing for raw AI output and negotiating feedback can be maintained.
- Defensive AI use (e.g., summarizing emails) may emerge, but it risks losing valuable human feedback and eroding openness in academia and open source.
- Long-term relationships allow calling out externalization and establishing norms, though this gives up on broader openness and efficiency.