A máquina pode gerar a surpresa, mas transformá-la em significado continua a ser uma tarefa nossa
For a long time, the idea that a machine could be creative seemed hard to take seriously. Today, however, anyone can ask an artificial intelligence system to write a poem, compose a song, or design a logo—and receive dozens of suggestions in just a few seconds. Therefore, it is no longer absurd to ask whether a machine can be creative. However, to determine whether a machine can be considered creative, we must first understand where to look for creativity. There is no consensus on this among either philosophers or AI researchers. In both fields, some evaluate creativity based on the resulting output, others focus on the process that produced it, and still others consider the system itself to be creative.
This distinction is important because a creative result does not necessarily imply that the process that gave rise to it is also creative, much less that the system can be considered a “creative agent.” The discussion, however, often shifts from one level to another. Some deny that AI possesses any creativity because it merely recombines materials produced by human beings. Others see in the texts and images it produces proof that the machine itself has become an author like us. Both positions jump to conclusions about the system too quickly. Before doing so, it is necessary to distinguish between the three levels, starting with the result.
The philosopher Margaret Boden argues that a creative idea must be new, surprising, and valuable. Generative systems already seem to satisfy the first two criteria with relative ease; they produce results whose exact form was not previously defined, and some of these results are also valuable to those who use them. Therefore, if we evaluate only the final product, we can speak of a creative result. But that does not yet mean that the process that generated it is “creative” or that the system itself can be considered an author.
To say that AI merely recombines materials produced by humans does not, in and of itself, resolve the issue. Human creation also builds upon rules, methods, and previous works. A sonnet follows a specific form, a Bach fugue unfolds according to musical rules, and a mathematical proof follows formal rules. As Berys Gaut asks, the question is whether the creative process is limited to repeating patterns or whether it uses them to open up new possibilities.
Finally, there is the system itself. Producing a creative result is not enough to make it a “creative author.” Philosopher Matthew Kieran distinguishes between the ability to produce something new and valuable and a more general form of creativity. In this sense, creating involves a purpose, mastery of a practice, and the ability to recognize the reasons that make one choice better than another. An author can explain what they sought to do, review their decisions, and take responsibility for the work. However, current AI systems do not initiate a project because something matters to them; they are given defined objectives and, as far as we know, do not have a biography, commitments, or an experience of the world from which creation acquires meaning. Ricardo Barroso Batista (2026) 1 It is precisely here that Boden’s third criterion (value) becomes essential to the discussion of creativity. A system can learn patterns of preference and estimate which of its responses will be best received by a given audience. But predicting a reaction is not the same as understanding why a proposal has value or why it deserves to be developed. To distinguish a truly new possibility from a mere repetition of a cliché, one must be familiar with the field and tradition in which one works. A user may obtain a correct and fluent text from AI without recognizing that it is “trite”; An experienced user can use the same tool to compare alternatives and find a solution they hadn’t yet considered. AI can shift the focus of creative work. When it’s possible to generate dozens of proposals in just a few seconds, producing “alternatives” becomes “easy.” The challenge then lies in choosing and filtering what’s worth keeping, developing, or discarding. AI is a “new tool” that can present “new possibilities” to users, but someone will have to “creatively” sift through them and take responsibility for what they decide to do with them. We can, therefore, recognize creativity in many results produced by AI without necessarily turning the system into a “creative person.” The tool can enhance human creativity, but that contribution is not enough to make it an “author.” It is important to know who recognizes the value of the result, who decides to keep it, and who is accountable for it. The machine may generate surprise, but transforming that surprise into meaning remains our task.
Ricardo Barroso Batista he is a professor and researcher at the Faculty of Philosophy and Social Sciences at the Universidade Católica Portuguesa
