Van Art Gallery: AI Installation

Designing an award-winning AI installation for the Vancouver Art Gallery so visitors' movement through the room became the artwork.

Problem

The gallery's head curators asked my team to design, develop and implement an installation for Imitation Game, their exhibit on AI and visual culture, which ran from March to October 2022. It had to gather visitor data without getting in anyone's way, and show it back in a form that was attractive and "accurate-ish".

01Research

Visitors would arrive with completely different relationships to AI, and the piece had to hold all of them at once.

I plotted attitudes to AI on an audience spectrum, from 'the user is in control' through to 'the AI is a companion'. Two personas sat at either end: Steve, a retired builder, and Ashley, a museum learning advocate with a data science background.

Audience spectrum matrix, plotting attitudes to AI from jaded to intrigued, and from AI novices to AI aficionados — Fig. 1The audience spectrum, from jaded to intrigued and from novice to aficionado. The whole piece grew out of it.

Persona: Steve, a retired builder of fifty-five — Fig. 2The two personas. If the piece worked for both, it would work for the room.

Persona: Ashley, a museum learning advocate with a data science background — Fig. 2The two personas. If the piece worked for both, it would work for the room.

02Concept

Each visitor became an anonymised, numbered, coloured path, drawn live on a nearby monitor.

The curators chose my idea: real-time human-detection software plotting movement and dwell time as it happened, so the piece doubled as artwork and research tool. The visitors became the artists.

Test 1: the line

Test 2: standing still

Fig. 3Early tests of the line, and of what it does when somebody stops moving.

03System

The line was a vine rather than a graph, so it grew.

I named the piece Creepers, and the name works both ways. It came from the feeling of being surveilled: cameras quietly following you round a room is a little creepy. And the lines climb and spread across the screen like a creeping plant.

I wrote its rules before anything was built. Colour was set by the hour of the day, dwell time rippled out as rings, and lines faded to 66%, then 34%, then out as each session's visitors left.

Design system board: line colours by hour of the day, dwell-time rings, and the logic for expiring lines — Fig. 4The rules, written before the build. 1Colour by the hour. 2A ring for each stretch of dwell time. 3Lines fading out as new visitors arrive.

04Room

The rules had to be explained through the piece itself, so the room was planned around it.

I mapped the visitor journey and exhibition layout from the personas, and wrote the didactic panel. The room held one screen and one line drawing, and was kept dark enough to let the drawing carry it.

Steve's visitor journey map, from pre-visit through the exhibit to telling his family about it — Fig. 5Steve's journey. His questions in the room ("We are data? How is that even possible?") are what the wall had to answer.

The installation mocked up on the gallery wall — Fig. 6The proposal: the drawing placed on a floor plan of the gallery, with a 'you are here' and footsteps leading to it.

Close-up proposal: the gallery floor plan on the wall, the drawing in the room it maps, a 'you are here' marker and footsteps on the floor — Fig. 6The proposal: the drawing placed on a floor plan of the gallery, with a 'you are here' and footsteps leading to it.

Creepers

Artificial Intelligence has edged its way into our daily lives, from predictive web searches at home to tracking visitor movements and displaying the visual output in galleries, as you see right here.

Along with every other visitor that has passed through the Salon des Refusés space, the path you took is now data. It was captured by AI-driven cameras that harness object detection to track movement. The input has been converted into a real-time, ever-evolving art piece.

Creepers’ output is completely dependent on how people move through the space. The only constant variables are the visual elements. Each line represents a visitor. Each colour represents the time of day they passed through and the rings indicate the amount of time someone spent in a particular area within the Salon des Refusés: 1 ring = 1 minute.

Fig. 7The didactic panel I wrote for the wall. It tells you that you're already in the piece before explaining how to read it.

05Exhibit

Every visitor was detected, numbered and given a colour. The number and the colour were all anybody became.

Creepers was meant to be understood by any visitor, and to look good enough to enjoy without the deeper meaning. The paths cross, loop back and double up the way people actually move.

Input: what it saw

Output: what it drew

Fig. 8One session, in and out: each id on the left becomes a line on the right.

A visitor watching the installation in the gallery — Fig. 9In the gallery: a visitor watching the drawing build (left), and the room in use (right).

Visitors in the exhibition space, each already carrying an id and a colour — Fig. 9In the gallery: a visitor watching the drawing build (left), and the room in use (right).

Results

~830 average visitors per day
across the run

8 months in the exhibit

1 Award: Gerri Sinclair for
Innovation in Digital Media

Awarded at the Centre for Digital Media.

What I learned

You can't design one explanation for a room full of people who arrive with completely different relationships to AI. The piece had to explain its own rules.