How smart is “smart” technology?

written by Andrea Wojcik
11/01/19

11th JANUARY 2019, MAASTRICHT – In December I attended a workshop organized by the Netherlands Graduate Research School of Science, Technology and Modern Culture (WTMC) on “smart” technologies. It was a stimulating three days that allowed for surprising connections and interesting questions related the ‘Making Clinical Sense’ research project. In particular, the lecture given by Merel Noorman, from the Tillburg Institue for Law, resonated with something I came across at the University for Development Studies in Tamale regarding learning physical examination skills—the potential bias of medical education material found online.

Noorman spoke about the role of science and technology studies (STS) amidst the current fascination and application of smart technologies. She drew on a number of examples to build her talk, but one that struck a chord with me was a Google image search of a three letter acronym—CEO. This isn’t the case anymore, but back in 2015, the first woman to appear in that image search was Barbie. Gender bias clearly made its way into Google’s image search algorithms.

Jumping over to medical education, Google is often a good friend to medical students. Need a quick physiological explanation? Google it on your smartphone. Don’t know how a condition may present clinically? Google it on your smartphone. Surely, having information at their fingertips has benefited medical students all over the world, but the story above does raise some questions about the limitations of immediate information.

For example, students were learning about cyanosis in class. According to a quick Google search: ‘Cyanosis refers to a bluish cast to the skin and mucous membranes…It’s usually caused by low oxygen levels in the red blood cells or problems getting oxygenated blood to your body’. So it is recognisable via a bluish tint. The class tutors explained that the tint is usually visible on the hands and feet and/or the lips, tongue and gums. The following image is of a Google image search for ‘cyanosis’ zoomed out at 50%. Anything striking?

Google image search for cyanosis, zoomed out 50%

In class, tutors informed students that a bluish tint would be more difficult to recognize in people with dark skin, which is the majority of their patient population. They also pointed out that a Google image search would not be helpful in generating images of how cyanosis presents in dark skinned people. Indeed, the image search above, zoomed out to 50%, came up with only a few presentations of cyanosis in a dark skinned person’s hands and feet and one image of a dark skinned persons lips, tongue or gums. (What might appear to be an extreme case on the bottom right is actually artwork.)

One of the biggest takeaways from the WTMC workshop was that “smart” technologies are sometimes quite dumb. Perhaps it comes as no surprise, but Google clearly doesn’t have all information one click away. For the medical students in Tamale, this isn’t per se a problem. They will come across presentations of cyanosis in dark skinned patients when they enter the clinic.

But as Noorman pointed out in her lecture, the role of STS is to ask the big picture questions. Beyond raising questions about who benefits from smart technology in medicine, then, we have to question the smart technology itself. Noorman made it clear that questioning how to eliminate bias in smart technology isn’t the only question we can ask. Besides, many STS scholars would argue that such a question is futile given that we, human beings, make the tech and feed the algorithms. Instead, we have to decide how to use smart technology. We need to learn what it can and can’t do. In medical education, this might mean emphasizing that Google can’t stand in for other important sources of learning, such as clinical experience, either of the students themselves or their teachers.

The workshop was an intimate affair with participants coming from a variety of backgrounds. We were a mix of medical specialists, young doctors, a few biochemists, tutors, administrative staff, and quantitative and qualitative researchers. While a few participants used the workshop as an opportunity to explore qualitative methods for the first time, most participants had conducted interviews in some fashion, whether to take a patient’s medical history or to generate research, and as such were able to draw on a variety of experiences to inform discussions.

John and I divided the workshop into roughly two parts. We began with a discussion of the potential value of qualitative methods for medical research. We provided some examples of research which has inspired us (e.g., that of Margaret Lock, Vinh-Kim Nguyen, and Caroline Bledsoe and Jennifer Johnson-Hanks), and John was able to draw on his experience of researching the history of nutrition in Ghana during his PhD to briefly highlight the shortcomings of national surveys and the possibilities of using mixed methods.

After this, we bowed out of the limelight and asked participants to conduct their own interviews in a role playing activity. They were divided in groups of three or four, and each group had an interviewer, an interviewee, and at least one observer. We instructed them to choose their own interview topics, invent characters, and play with format (e.g., structured or semi-structured); the goal was not per se to conduct “the best” interview, but to experiment with the setup.

Participants identified difficult aspects of conducting interviews

Upon reconvening, the groups drew on their experiences of doing the activity to inform the closing discussion. Together we touched on a variety of topics ranging from some of the difficulties and ethical dilemmas associated with conducting interviews to general tips for good practice, which John recorded for distribution later via email.

We thank those who attended the workshop and shared their experiences of interviewing with us as well as SMHS for allowing us to facilitate the workshop. We must as well extend our gratitude to the European Research Council for funding the ‘Making Clinical Sense’ research project, of which this workshop is a part.

Hyperbolic modelling

Charles the plastic skeleton was a constant companion of the students in their skills lessons the last five weeks. Even when it seemed he managed to escape to the sidelines, he was pulled front and center to make a point for the class’ benefit. One instructor would temporarily neglect the living, breathing student volunteer who sat on an examination bed in front of the class and instead rely on Charles for demonstrating the limitations of human limbs. The instructor would, for instance, extend Charles’s leg at the hip in the direction of his spine to the point that the leg was perpendicular to his upright back, a move that would have been nothing short of grotesque in someone of flesh and blood. The purely skeletal nature of Charles’s dangling limbs, held together by wire pins, allowed for something the healthy student volunteer did not, a kind of hyperbolic modelling which ironically highlighted the normal degrees of human movement and invited an imagination of the muscles and ligaments needed to adequately limit Charles’s movement.

Charles the skeleton in his sidelined position

Embodied modelling

At times, however, even the models were too limited in their movement, and instructors would default back to the body, but not necessarily to the part of the body under scrutiny. For example, before going through the protocol for examining the knee, an instructor used a model of the knee to zoom in on two ligaments nestled in the hollow between the femur and the tibia. These ligaments prevent the tibia from dislocating. Then the instructor tried to demonstrate what would happen if the ligaments were torn, holding the floating femur with one hand and pulling at the tibia with the other (see my reenactment below). After a few attempts, the instructor put the model down, made fists, and placed them against each other with one forearm running towards the floor and the other towards the ceiling, creating a model of the knee using hands and forearms. This model was easily able to demonstrate the abnormal movement of a dislocated tibia resulting from torn ligaments in the knee.

During the musculoskeletal physical examination block, instructors creatively utilized models and modelling to teach students about bodily movement. This often took the form of using a model to exaggerate an abnormal degree of human movement to ironically invoke the normal. At other times, the models were too securely and representatively constructed to deviate from normalcy for the purposes of education. When this was the case, instructors created models on the spot, using their own body as the resource.

Sensory ethnography is becoming an increasingly popular method for exploring taken-for-granted practices that are otherwise difficult to articulate. Sensory ethnographers are often asked to attune to their own learning, to learn with, not about, others. Much of the discussion, however, is focused on the solo researcher, and there is limited attention given to collaborative studies of the sensory. In a quasi- “proof of concept study”, our team – Rachel Allison, Anna Harris, and Andrea Wojcik – sets out to experiment with different digital methods of elicitation and notation which imaginatively attend to sensory learning – namely drawing, photography, and video – while taking another example of a sensory skill that demands finely tuned technique – making omelettes.

Cooking an omelette from a video

Because our interest in the medical world is the role of pedagogical technologies, we will document others learning through the use of different technological arrangements: video (Julia Child’s The French Chef); written recipe (M.F.K Fisher’s How to Cook a Wolf); and apprenticeship (under the guidance of a chef). In doing so, we mimic the arrangements we expect to encounter in our individual fieldsites. Working from these arrangements also enables us to cross disciplinary boundaries and collaborate with media and art historians to understand more about what it means to experimentally re-enact and reconstruct recipes.

To recap, our starting point is the awareness that our methods don’t describe practices but rather help create them. Through ethnographic experimentation, we look for insights from other disciplines – such as media and art history – to be able to understand how we can create practices which inform us about difficult to articulate sensory experiences of learning to cook an omelette. This experiment should help us direct our use of audio-visual methods as a collaborative team studying doctors’ learning of sensory skills across three fieldsites.

We look forward to conducting this experiment in the summer 2017, and will report back on our experiences later in the year.

filter entries
Event

Time for tinkering

by Anna Harris
February 21, 2019
Event

How smart is “smart” technology?

by Andrea Wojcik
January 11, 2019
Materials

Tools of the teaching trade

by Sally Wyatt
January 10, 2019