Critical Objects Analysis

Final Major Project / 07

September 30, 2021

Luchen Peng, Tiana Robison, Jinsong (Sylvester) Liu



A synthesis framework


Thanks to all the generous participants, we have collected roughly ten critical design ideas of data bias with continuous workshops across the summer break. While a systematic analysis is crucial to understand them better and prepare for the next stage, it also stays tricky and requires creativity.

I agree with Malpass (2017) on a more design-centric analysis in critical design practice, instead of an art-oriented view. Our group synthesis some of the questions and the Artefact Analysis to the framework below, which I will use to demonstrate two examples.




Example from Data Scientist (Hassan)



This is an intelligent supervision bracelet called "your right-hand man", which could correct your behavior by constantly slapping your hands to help you get a healthy body.

Design: The device includes a colourful, flexible wrist band (LGBTS), a circular chip and a movable hand figure on top.

Interaction: When users want to buy products (e.g. donuts) that affect their “health credit”, the “hand” will slap on the back of their hand, giving them pain to prevent purchasing.

Social Interpretation: It appropriates the concept of financial credit into health, then exaggerates the mental struggle into physical punishment.

Audience: people who wants to stay a healthy life habit and want some reminding and supervise, or has long term disease like diabetes. It is owned by the consumer.

Context: When a person’s “health history” will become a significant impact in their credit, they will use it in food markets.

Theme of critique: It uses data of food ingredients, brands, personal medical history. The design tackle gender and racial bias in health. How do we define health? Do we want to sacrifice autonomy to get “the absolute health”?

Design by Hassan.


Example from non-expert (Jasper)


Because the training data primarily comes from the lighter-skinned population, the AI system detects skin cancer could potentially misdiagnose those with darker skin. This scanning machine would create every combination and test each one, no matter the human’s race or gender.

Design: The machine includes a scanner (metallic cylindrical cage) and serval spinning sheets (transparent material) that generate different gender combinations. A marginalised race person (coloured humanoid) stands in the middle to test whether the algorithm is biased.

Interaction: When step inside the cage, it produces several combinations of colour and gender. of the patient. The scanner scans everyone of these versions to ensure fairness.

Social Interpretation: It appropriates the concept of financial credit into health, then exaggerates the mental struggle into physical punishment.

Audience: doctors would use this on all their patients who are seeking out to test for skin cancer to make sure everyone is given equal diagnosis.

Context: This would be used during a AI diagnosis of skin cancer in a hospital.

Theme of critique:  It uses image data of previous skin cancer cases and images of the patient’s skin. The design tries to detect and solve racial and gender bias when AI meets medical treatment. The algorithm doesn’t know which one you truly are and has to diagnose all versions of you.

Design by Japser.



Findings


It is worth mention that not all the workshops went smoothly. In some cases, participants found it struggling to understand or don’t know what to do. Some occasionally dive into digital privacy since it’s more personal related. Experts from a computational background obviously digest the subject easier and deeper.

Moreover, we discovered that while some of us might hold an optimistic view of AI, asking them to be satirical isn’t revealing their genuine attitude. Therefore, the discussion part should be emphasised more instead of “just make it real”.  

We’ve been iterating this analysis method, mostly what questions to ask, and applied it after collecting ten design ideas. While it is beneficial to investigate the design in-depth, we think it could be better to involve it throughout the creation. Our next step is to refine some of the design and try prototyping.


References


  • Malpass, M. (2017). Critical Design in Context: History, Theory, and Practices. London Oxford New York Bloomsbury Academic, an Imprint of Bloomsbury Publishing Plc.