A question about artificial intelligence in music usually arrives already shaped. Is this song allowed on the platform? Who owns this melody? Was this voice used with permission? Each picks a decider, a moment and a kind of answer before any evidence is read. This study looks at a set of public records that try to hold the question open at the stage before that choice is made, and asks what they settle and what they leave unsettled.
The records come from Dionyas, the music studio of the RISCON family. Dionyas does not publish judgments. Its list of 1,539 inquiries describes itself as editorial question design, and it states that it has generated 0 judgments. Each inquiry is a question with a stated scope, a short list of the material it would need, a boundary that says what the inquiry may not claim, and a state such as "awaiting judgment". The list's own type is editorial question design, and it counts no generated judgment. What these records offer, then, is not a finding about music but a design for finding something out. That is exactly the stage that most writing on decisions passes over.
The research question and the contribution
The research question is this: how does an inquiry about AI in music frame a decision before any judgment is made, that is, what does it ask, what does it cover, who would decide and when, and what would change it?
The study reads sixteen inquiries chosen by one rule, set out in the method section, and it makes two contributions. The first is an account of a particular form of problem formulation: one that is written down, published and left unanswered. The literature on formulation mostly studies it inside a person or a team, as a step on the way to a solution. Here it stands alone, as a public object, and its choices can be read line by line. The second contribution concerns the debate on AI in music. The inquiries take the contested questions of that debate, about authorship, consent and the material a system is given, and recast almost all of them as questions about what record should exist. That move has consequences, and the study tries to name them without deciding any of the questions the records leave open.
Problem formulation and problem finding
A long line of work holds that how a problem is put shapes what can be done with it. Roger J. Volkema, in Problem Formulation in Planning and Design (1983), treats the role of problem formulation in planning and design, including the problematic, physiological, psychological and environmental factors that can affect it, and he sorts formulation heuristics into two families: problem reduction and problem expansion. In his experiments a heuristic of expansion, which he calls Problem-Purpose Expansion, may have helped people generate ideas, particularly on problems outside their own field, while simply urging people to take formulation seriously produced little measurable effect. The distinction between reducing and expanding a problem is useful here, because the Dionyas inquiries do both, and in a pattern.
Mark A. Runco and Jill E. Nemiro, in Problem Finding, Creativity, and Giftedness (1994), review research showing that problem finding is distinct from problem solving, and they describe it as a family of related skills: identifying a problem, defining it, expressing it and constructing it. The four words matter for this study. A published inquiry is a problem expressed; it may or may not be well identified, defined or constructed, and the records let these steps be told apart.
Markus Baer, Kurt T. Dirks and Jackson A. Nickerson, in Microfoundations of Strategic Problem Formulation (2013), start from the point that a problem must be formulated before a strategy can address it. They predict a core set of impediments that arise when complex, ill-structured problems are taken up by heterogeneous teams; these impediments narrow formulation and so limit the search for solutions. They treat the impediments as design goals for a structured process, and give collaborative structured inquiry as one example. This study reads a studio that writes its inquiries from a shared table of domains, as Dionyas does, as having built something of that kind: a mechanism that makes every formulation pass through the same fields. The fit is partial, since Baer and his co-authors write about teams forming strategy, not about a public catalogue.
David Dery, in Agenda Setting and Problem Definition (2000), separates two things that are often run together. Dery, through the case of one social protest in Israel, the 1990 tents movement, holds that to legitimise an issue is not the same as to legitimise the demands attached to it. Dery calls attention to the power of initial definitions of problems. This study reads that as a reason the design stage deserves study: whoever sets the first definition of a question about AI in music may set much of what will count as an answer.
The debate on AI in music
The second body of work concerns the questions themselves. Bob L. T. Sturm, Maria Iglesias and others, in Artificial Intelligence and Music: Open Questions of Copyright Law and Engineering Praxis (2019), look at AI applied to music from two sides, copyright law and engineering practice, through one specific application of AI in music creation. By their own account most of the questions they collect are open, because the answers are not yet clear. This study reads the paper as, in effect, a list of inquiries.
Andre Holzapfel, Bob L. Sturm and Mark Coeckelbergh, in Ethical Dimensions of Music Information Retrieval Technology (2018), argue that such technology is not value-neutral but shaped by design choices, so that it has ethically relevant effects which stay invisible unless one asks how it relates to wider society. Their argument concerns systems, not inquiries, but this study borrows its form: a question about technology is also designed, and its design choices also have effects that are easy to miss.
Eric Drott, in Copyright, Compensation, and Commons in the Music AI Industry (2020), observes that legal debate has mostly asked who should hold rights over what an AI puts out, and has largely overlooked what is put into it, the music that makes up the training set. He argues that copyright, built on individual rights, is not adequate to the questions of distributive justice these systems raise, and he sketches commons-based remedies. Fabio Morreale, in Where Does the Buck Stop? Ethical and Political Issues with AI in Music Creation (2021), names two urgent issues of commercial AI-generated music: a growing surplus of artists with a falling cost of creative labour, and the tacit acceptance of neo-colonial practices that exploit existing music and listeners' preferences. Both papers move the question away from the single work and the single owner. The test for the Dionyas inquiries is whether they make that move too.
Two further works give the study its terms for what the inquiries keep reaching for. Amy Brand, Liz Allen and others, in Beyond Authorship: Attribution, Contribution, Collaboration, and Credit (2015), hold that, as the number of authors on scientific publications grows, an ordered list of names is inadequate for credit, and report a contributor role taxonomy produced by a multi-stakeholder group. Mark Sheehan, in Can Broad Consent Be Informed Consent? (2011), argues for biobanks that broad consent can be informed consent, and that the differences among kinds of consent are differences among the kinds of choice a person makes. Neither is about music; this study reads both as concerning what a record of contribution or permission can hold.
What follows
The first chapter gives the data and the method, then reads the inquiries in two sections: how they turn questions of rights into questions of records, and where they place time, the decider and what would change an inquiry. An appendix gives the sixteen records; the conclusion returns to the two debates.
The contribution, stated at the start so that it can be checked at the end, is this. The Dionyas inquiries show problem formulation as a public, standardised act; they expand the problem toward documentation and contract the claims an answer may make; and in doing so they keep the decider and the deadline out of the question, which is both their discipline and their limit.