My AI Usage Guidelines

This is a place for me to document for myself my own guidelines for using AI. These are written with the core assumption that doing something yourself slowly is better than having an AI do it quickly. This ultimately means that I prefer to not use AI at all, however sometimes the speed of task completion and the fear of missing a deadline takes precedence over my own value of my own labor, creativity, and learning. That is a particularly difficult thing for me to admit, and it is my hope that somehow codifying these guidelines will make me more intentional in my use of AI.

To clarify, by AI I mean the usage of large language models (LLM), either through chat or agentic interfaces. These guidelines don't strictly apply to non LLM-based tools.

The Guidelines

There are currently three guidelines which, if not adhered to, constitute an irresponsible usage of AI.

  1. The output of the system must be able to be validated against an organic test suite.
  2. The input data to the system must be well understood by me.
  3. The outputted object must be clearly labeled as created using AI.

These guidelines are quite generalizable, but their power comes from their specific applications. I tried to avoid caveats or edge-cases which I might use to avoid proper usage. I will now expand on each of these in more detail.

Organic?

I use the adjective organic to describe any process or object created using only human-written tools. Organic, as in deriving from living organisms, but also of organs, of a living, breathing, decaying body. At the same time, organic as in free from non animal or plant-derived fertilizer: arising from an embodied object, in as so far as that body will one day fail.

This is an incredibly useful term when describing the usage or avoidance of AI systems. It allows me to describe an assemblage of sources without having to rely on -- but also encompassing -- "human" as an originating factor. I would consider the use of the traditional internet "organic." But the real strength of the term comes from explicitly connecting AI systems and tools to their material conditions of resource extraction and power consumption. In its recent usage to describe commodities, "organic" quickly became a political and economic battleground to distract the average person from the mechanical processes of extraction which many foods or products labeled as "organic" rely on. It is my intention to take advantage of this ambiguity rather than to dismiss it: the so-called "organic" sources such as code documentation or books may themselves be a part of processes of extraction and appropriation, and soon enough will themselves contain obfuscated, undeclared uses of AI. An organic object (or in the case of food, product) is only organic in relation to the non-organic object next to it. Organic chicken breast is organic because it qualifies a list of requirements, whereas the adjacent, cheaper non-organic chicken breast does not. And today, many families cannot afford organic food.

I don't want to give the impression that the above paragraph is a satisfactory discussion of the benefits or pitfalls of using the term to mean "non-AI generated" (much less a discussion of the complicated question food availability and pricing, one that is interlaced with larger discussions of class and race). However, it does serve as a sufficient provocation and justification for its use, which is all I need for the purposes of this post.

1. The output of the system must be able to be validated against a test suite.

2. The input data to the system must be well understood by me.

3. The outputted object must be clearly labeled as created using AI.