Keynotes

Re-enchanting the Pixel through Digital Kinship and Electric Spirit Berries

For decades, computer graphics has operated under a “bad habitus” of applying colonial logic and the Cartesian grid, which flattens complex, relational worldviews into discrete, extractive data points. This talk traces a decades-long journey of computational resistance, beginning with the pixel’s ancient origins in celestial “constellations” and culminating in a practice of Indigitalization.

By reframing the pixel as a manidoomin (spirit berry), we move beyond Western formalist models of point, line, and plane toward a framework of Digital Kinship. This talk demonstrates how intentionally “breaking” programmatic efficiency by replacing linear loops with cyclical spirals and addressable grids with relational structures can transform understandings of digital visual representations into a site for Indigenous and ancestral survivavnce and cultural continuity.

We will explore how Indigenous epistemologies can reorient the future of multimodal research, challenging the graphics community to look beyond optimization. How might our systems change if we treated technology not as a neutral tool but as a relative? This session offers a path toward “re-enchanting” our digital canvases, asserting that the future of the interface lies in its capacity to sustain rather than extract and to remember rather than erase.

Cues, Not Commands: Designing for Intent Beyond the Prompt

Imagine you are collaborating with a colleague on a design. As you discuss layout, colour, and content, neither of you speaks in precise instructions. Instead, you sketch loosely on a whiteboard, gesture through possible arrangements, and articulate half-formed ideas as they come to mind, sometimes leaving things deliberately ambiguous, sometimes clearly articulating constraints. The explanation is fragmented, evolving, and often imprecise, yet punctuated by a few specific, fixed, non-negotiable elements. When your colleague pauses and asks, “you see what I mean…?“, you nod. From these fragments, you’ve grasped the intent, the big picture, and can move the idea forward.

This scenario illustrates a distinctive human ability: we are remarkably good at making sense of incomplete, multimodal, and approximate information. We infer intent, fill in gaps, and extrapolate beyond what is explicitly said, while respecting constraints and staying within an intended direction. With the rise of generative AI, the vision of interacting with computers in this way has become more tangible. From a loosely expressed idea, even a ramble, systems can now generate images, text, and structure. Interaction is increasingly reframed as conversation.

And yet, this shift exposes a tension. Traditional computing systems have long relied on instruction-driven interaction: rigid, precise, and controllable, but overly literal, doing exactly what is specified and nothing more. AI-powered conversational input, on the other hand, creates the illusion that intent can be directly expressed and faithfully realized. In practice, we are still forced to spell out details, by expanding and refining prompts through low-level textual instructions to maintain control over the things we had in mind.

In this talk, I argue that we are not lacking better prompts; we are missing forms of interaction that offer high control while embracing underspecification and ambiguity. The answer is not to shift traditional input toward open-ended conversational modes, but to rethink the nature of interaction itself, beyond operation-level instructions. Operation-level instructions give us precision and control, but they capture only part of what we mean. The real challenge is to elevate them into forms that express both exact specifications and intent.

Focusing on creative domains such as drawing, sketching, and writing, I will show how intent can be expressed through augmented, non-conversational inputs and tools that preserve strong authorial control while leveraging the automation and generative capabilities of modern computing. These approaches move push interaction beyond instruction-following, towards intent, embracing emergence rather than merely executing deterministic commands for computationally-supported creativity.