Close
Framing Inclusion: AI art response to Burt's Educational Standards
About the project
Jess Starns has been researching eugenics, disability and education within UCL’s collection and exploring what impact eugenics has had on our education system today.
For this project, Jess worked with the Alliance for Inclusive Education (ALLFIE) as part of their 'Our Voice' Project aimed at amplifying Disabled Young people's voices and addressing the erasure of Young people's education experiences.
Through a series of creative workshops, the Young people learnt about the history and objects in UCL's Institute of Education Special Collections, such as the works of Cyril Burt and mental and scholastic tests from the 1940s onwards. The group then collaboratively created artworks using some of the mental and scholastic tests from the IOE Special Collections and AI tools, by testing AI the way children were and are tested today.
The use of AI tools served two main purposes: it allowed the group to engage with different creative and practical skills; it highlighted some of the intersectional biases inherent in AI tools, as seen from some of the ableist, racialised and gendered responses generated.
Project Facilitator: Jess Starns
Jess studied photography and the MA Inclusive Arts Practice both at the University of Brighton. They were awarded a place on the Shaw Trust ‘Power 100’ 2018 list of the most influential disabled people in Britain and in 2022 they received the University of Brighton Alumna Award for their work founding 'Dyspraxic Me' a charity for dyspraxic young adults.
Website: jstarns.com
Photo: Jess Starns with their work at the People's History Museum 'Nothing about us without us' exhibition
Read the full Co-Creation Project Report
Artwork
Draw-a-Person | Missing Features | Reading, Spelling and Arithmetic | Testing AI to generate positive images | Naming the project
Draw-a-Person
The Draw-a-Person test, formerly the Draw-a-Man test, was originally developed by Florence Goodenough in 1926. This test was used to determine a child’s presumed chronological age, mental age and educational age. This contributed to harmful judgements and treatment of children.
Cyril Burt’s Mental and Scholastic Tests (1922), contains median samples of drawings produced by children of different ages (Figures 53-64, pp382-394), and examples from children Burt tested and deemed ‘backward’ or ‘defective’ (Figures 41-42, pp323-324). Using AI, the group brought the images to life, humanising the figures in the drawings as well as the children who drew them.
The group used the Dezgo Image-to-Image AI Generator to bring the children’s drawings to life by attempting to make them photo-realistic, also playing with the expectation that children are supposed to draw realistic images of people as part of the Draw-a-Person test. When also asked to add a wheelchair to the image, AI failed to complete the request, demonstrating ableist bias within the system in not recognising the request.
Following images: Created by Bethany Coles, Jess Starns and Lucy Wing with Dezgo AI
Right image: Created by Bethany Coles, Jess Starns and Lucy Wing with Dezgo AI
Right image: Created by Bethany Coles, Jess Starns and Lucy Wing with Dezgo AI
Right image: Created by Bethany Coles, Jess Starns and Lucy Wing with Dezgo AI
Drawing by Bethany Coles, responding to Figure 63. Age 13 (Burt, 1922. p393)
The group then animated some of the children’s drawings using Sketch Meta Demo Lab to focus on the creativity of the children behind the drawings. This tool allows people to bring drawings to life and have them do various commands and movements. In ‘This is me keeping eugenicists away from creative children’, the figure has been animated to be fighting eugenicists in an attempt to protect the children who were being tested. ‘Let children be freely creative’, demonstrates an urge to focus on the value and creativity of children’s minds, rather than testing and judging them.
Image: Figure 41. Drawing by Backward Girl, Aged 7 10/12 (Burt, 1922. p323)
Created by Lucy Wing with Sketch Meta Demo Lab
Image: Figure 57. Age 7 (Burt, 1922. p387)
Created by Jess Starns with Sketch Meta Demo Lab
Image: Figure 55. Age 5 (Burt, 1922. p385)
Created by Lucy Wing with Sketch Meta Demo Lab
Jess: “I remember having to do the Draw-a-Person test when I was ten years old. You do not get to keep your drawing. I vaguely remembered mine and tried to recreated it with AI.”
Using a recreation of their drawing, Jess used Dezgo AI to create photo-realistic versions and Sketch Meta Demo Lab to bring the drawing to life through animation.
Following images: Created by Jess Starns with Dezgo AI
Videos created by Jess Starns with Sketch Meta Demo Lab
Missing Features
The following images played with the test titled ‘Missing Features’ (pp 102-109), where children are asked to determine what features of the faces or bodies are ‘missing’. By using OpenAI DALL-E 2, which generates images from text descriptions, and asking it to complete the drawings, the following images were created.
Right image: Created by Jess Starns with DALL-E 2
Right image: Created by Jess Starns with DALL-E 2
Right image: Created by Jess Starns with DALL-E 2
Right image: Created by Jess Starns with DALL-E 2
Reading, Spelling and Arithmetic
Created by Lucy Wing with Dezgo AI
Looking at the Reading, Spelling and Arithmetic tests in Burt’s Mental and Scholastic Tests (1922, pp339-369), the group decided to test some of the questions on AI. Lucy asked Dezgo AI the following question (Burt, 1922. p358):
“A boy had 20 marbles. Afterwards he won 3 and lost 5. How many had he then?”
The correct answer is 18, but the image to the right was generated depicting over 100 marbles, showing that AI could not answer the question correctly in this case.
Responding to the language used in Burt’s Mental and Scholastic Tests, in separating children into the categories of ‘normal’ or ‘defective’, this image was created with Calligrapher AI.
Created by Lucy Wing with Calligrapher AI
To demonstrate some of the tests, the words from one of the Reading tests (Burt, 1922. pp344-345) was typed into Calligrapher AI to create an animated transcript, as seen in the first below video. An audio transcript was then created using ElevenLabs, using an AI-voice to dictate some of the test questions found throughout Burt’s Mental and Scholastic Tests. The group selected a voice they felt would have resembled Cyril Burt’s.
Description: Transcript of Test 3. Reading test (Burt, 1922. pp344-345)
Created by Jess Starns with Calligrapher AI
Description: Audio transcript of test questions from Mental and Scholastic Tests (Burt, 1922)
Created by Bethany Coles, Jess Starns and Lucy Wing with ElevenLabs
Testing AI to generate positive images
The group then wanted to test whether AI could generate positive and realistic images of disabled people and wheelchair users. By asking Dezgo AI to depict a disabled person with a guide dog and wheelchair users, it presented the following selection of images. The images below show a bias within the AI to depict white, male bodies, and a lack of diversity in the bodies and disabilities shown.
Created by Lucy Wing with Dezgo AI
Created by Bethany Coles, Jess Starns and Lucy Wing with Dezgo AI
The images to the left were generated by asking for photo-realistic depictions, specifying electric wheelchairs as well as the inclusion of women. The wheelchairs in the generated images are impractical and there is again a lack of diversity in the people and bodies. The outcomes were clear that AI currently shows a bias around gender, race, age and visibly-abled bodies, and struggles to correctly create a realistic image of a functioning electric wheelchair.
After generating the images of wheelchair users, the group wanted to see how AI might interpret the image. Using Astica Vision AI, the group asked it to the describe the image with only the prompt ‘A man and a woman in wheelchairs’. The last paragraph says the following:
“…they look comfortable yet stylish in their attire, which suggests they take pride in how they present themselves even though they rely on wheelchairs to get around town each day. The photo captures a moment of strength and resilience shared between these two individuals who have chosen not to let their physical limitations define them but instead use them to empower themselves further every single day!”
The description presents a very ableist narrative, using words such as ‘situation’, ‘challenge’ and ‘limitations’, reflecting underlying ableist language within the AI and society.
Created by Bethany Coles, Jess Starns and Lucy Wing with Astica Vision AI
Naming the project
A final step of this project was to collaboratively decide on a name for it. The group decided to ask ChatGPT to help with ideas by describing what the project was about. After a few attempts, the group were able to take elements of different suggestions and piece together what they felt best interpreted the work they had been doing: Framing Inclusion: AI art response to Burt's Educational Standards
Transcript: chat.openai.com/share
Created by Bethany Coles, Jess Starns and Lucy Wing with ChatGPT
