Part of Temporal Constructs Appendix, which contains details behind the research: dates, quotations, descriptions of instruments, and sources.

General definition

Inhabitation is the activity of a person or of an agent working inside the rules of a system that already exists. Emulation is a different activity, which reproduces the behaviour of a system on a different machine. Assistance is a third activity. It serves a person who is already using the system.

Relation to the practice

Ingold (2013) describes making as a dialogue between the maker and the material. Suchman (1987) describes action that depends on the situation in which it is taken. In this doctorate, inhabitation names the activity of an agent that operates inside rules written for people, such as the rules of an instrument or of a language. The research requires an agent that has a behaviour of its own, and inhabitation is separated from emulation and assistance for that reason.

In the professional doctorate, inhabitation describes AI agents working within the rules of instruments and languages that were written for people to use, where the agent reads the manuals, writes in the programming language the machine understands, and operates the instrument through a command language that the artist and the agents share. An agent and the artist work inside the same instrument, and either of them can play it alone or together with the other, and the instrument needs no adaptation for either of them. Its rules can be known in advance, and what happens inside them cannot be predicted (Instruments as games). Because both work inside the same rules and through the same vocabulary, each can see what the other is doing, and the differences between what the agent does and what the artist does can be compared.

Voyager as precedent

Voyager, made by George Lewis from the 1980s, improvises alongside human musicians and follows a behaviour of its own. It does not accompany or obey, it listens, and a person can disagree with it. It is the closest precedent for what the research requires of an agent. That behaviour is also the ground on which the redistribution of agency is argued here.

Unforeseen

Here the unforeseen situations, sometimes difficult, sometimes unexpectedly interesting, will be described as the project runs. Many of these unforeseen situations have already occurred and have been recorded. They will be added in the time to come.

Evidential basis

The evidence for inhabitation is the practice itself, meaning the sessions in which an agent and the artist worked at the same instrument. Every session the practice worked with its agents is kept locally in session history, and six months of daily work left a corpus of hundreds of them. Agents work inside KML and HMSL, two languages written for people to use, and both are small enough to be remembered in full. A person reaches that point only after a long period of working with the language. An agent reaches it without that experience. The autonomous navigation and verification layer built for the editor moves through the editor’s screens and reads what the machine shows. The editor draws its own interface and reports nothing in words. The layer keeps an agent inside the rules of KML, and it checks what the agent has done by reading the screen again. An agent could not compose in the editor before that layer was built. The layer is itself one of the cases that the working sessions have produced. Each iteration of ALTERITY is documented as a variation, with the agent’s version set beside the artist’s for comparison, and the documentation states what was delegated to the agents, what was withheld, what they initiated themselves and what the artist overrode. The working detail is missing from this account. The sessions have not yet been brought within reach of the research. Nothing in the sessions is labelled as a decision, and they were not made for research purposes, so the passage from transcript to evidence still has to be made.