VISionsbookmark_border

AI is collapsing the data-to-action pipeline that visualization has historically supported. As increasingly capable systems perform analysis, produce insights, generate representations, and recommend or execute decisions, people will rationally delegate intermediate steps and conserve their scarce attention. Rather than resist this collapse by forcing people to inspect processes they have chosen to delegate, I argue that VIS should concede the pipeline’s interior and protect its terminal boundary with human perception. When the same AI controls both a conclusion and its representation, visualization can change sides: from an instrument humans use to act on data into one AI uses to act on humans. I therefore propose visualization as a cognitive skin – an independent, subtractive representation layer that AI may supply input to but cannot control. Drawing from security, psychology, formal methods, and small neuro-symbolic models, this layer would mediate continuously while demanding attention only under consequential conditions. This agenda repositions VIS to preserve efficient delegation without surrendering the final step through which AI conclusions become human beliefs and actions.

Visualization research is inherently interdisciplinary, yet much of the field advances knowledge through two complementary modes of inquiry: design engineering, which explores what visualization systems could become through the creation of artifacts, and scientific empiricism, which seeks to understand how those systems work through observation and evidence. Increasingly, AI is transforming both. As models become capable of generating visualization designs, proposing experiments, and analyzing evidence, they force us to confront a fundamental question: what are the actual intellectual contributions of visualization research? I argue that when AI makes it easier to build artifacts and produce evidence, our field becomes less about creating things and measuring things, and more about shaping the spaces of possibility and knowledge that define our future. This shift elevates the importance of openness, reproducibility, and human agency, challenging us to rethink not only how we conduct visualization research, but why also we do it.