Beyond Data-Driven Aesthetics: Designing for Computational Visibility

Framing the problem

Haridis' exhibition traces historical ideas of aesthetic judgment and applies them to contemporary computational systems, arguing that design can make algorithms legible. For businesses deploying AI, this reframes transparency not only as a compliance checkbox but as a design challenge: how do you make complex models intelligible, meaningful, and trustworthy to users and stakeholders?

Significance for product and trust

Visible computation-clear visualizations, interactive explanations, and materially faithful metaphors-bridges the gap between technical complexity and user comprehension. This reduces user error, increases adoption, and improves accountability. When stakeholders can see and test the behavior of systems, it becomes easier to detect bias, question assumptions, and align outputs with organizational values.

Practical takeaways for leaders

Invest in cross-disciplinary teams (designers, cognitive scientists, engineers) to build explanation layers into products. Prioritize interface patterns that surface provenance, confidence, and trade-offs rather than opaque scores. Use exhibitions, workshops, and prototypes as stakeholder education tools to demystify models and gather feedback early.

Operationalizing interpretability

Adopt interpretability metrics and monitoring, and instrument systems to collect human feedback. Treat visual explanations as living artifacts: iterate them as models and use cases evolve. Finally, consider design-led initiatives as strategic assets for regulatory readiness and customer trust: when users can see what a system does and why, organisations reduce friction and liability while strengthening user relationships.