The REAL Card Deck: Thinking Differently About Algorithmic Systems

Text: Tuukka Lehtiniemi 

Our Reimagine ADM project has published a concrete tool to draw attention to human values in the design and development of algorithmic systems. The REAL card deck is designed to provoke and inspire individual and collective thinking about algorithmic systems. It is meant for experts, critics, developers, analysts, consultants, students, and researchers. Using the deck involves creative engagement in collaborative and multidisciplinary endeavours, and highlights crucial human aspects that are often overlooked.

The REAL card deck.
The REAL card deck.

The deck was inspired by our dissatisfaction with mainstream approaches to societal aspects of algorithmic systems. Well-trodden AI ethics discussions tend to focus on abstract principles that most would subscribe to, yet they offer limited practical guidance on how to proceed. There’s a need to build alternative approaches to how algorithmic systems shape our lives, society, and potential futures. And one thing that is needed for doing this is a new vocabulary.

Questions to probe keywords

The REAL (REimagining ALgorithmic futures) card deck is built around seven keywords: thresholdsteeringcodabilityreplacementresponsibilityrepair, and friction. These keywords emerged when we identified recurring themes in cutting-edge research on societal aspects of algorithmic systems. They are not intended as a definite or final list of what’s important, but rather a beginning for a new vocabulary, a sample of important concepts that deserve more attention.

In the card deck, each keyword has a set of eight question cards associated with it. The questions are designed to prompt reflection and dialogue on algorithmic systems. The aim is to stimulate  thinking and­ conversations about values embedded in such systems, as well as their social and technological implications.

Let’s examine one of the keywords as an example. Threshold signifies limits or boundaries, for example numeric or binary values that trigger responses from the algorithmic system. Threshold might define what is considered as a good enough match in a police facial recognition system, or in a dating app. It might determine when the conditions are met to automatically ban an account from a social media platform, or to flag a potential case of tax fraud or a risk case for child abuse.

These thresholds are the systems’ features, but they also reflect the aims and interests of the system’s stakeholders. The REAL cards pose questions about thresholds: Who has a stake in the threshold? Who defines it, and can it be redefined? What gets lost when complex things like human social behaviour are narrowed down to either-or thresholds? These questions guide those using the cards to think about algorithmic systems in terms of thresholds contained within them. Furthermore, the cards will invite their users to analyse thresholds alongside other keywords. Thresholds might, for instance, be analysed in combination with codability, which indicates whether or not a phenomenon, behaviour, process or experience can be translated into code; or steering, which describes situations when an algorithmic system guides its uses and users towards specific goals.

Together, the keywords and their questions create a powerful toolkit to assess algorithmic systems and their implications in our world.

Using the cards

The cards come as a physical deck. The digital version of the deck includes an overview of the cards, a booklet describing how to use the cards, a printable PDF file, and instructions on how to prepare the deck with an office printer.

Preparing the DIY printed deck.
Preparing the DIY printed deck.

There is no right or wrong way to use the cards. Similarly to Tarot cards, REAL cards provide a frame within which to reflect and look closely into a specific case or example of an algorithmic system. The REAL cards help their users to delve into the case, focus on certain aspects of it, ask questions about it, and discuss, debate, and deliberate. Depending on the goal, the cards allow comparisons across cases or alternative design choices. They aid in crafting stories about futures in relation to algorithmic systems, refining research questions or approaches, or focusing on observations or interview materials. The goal is reached when the discussion is exhausted or there’s enough thinking to base the next actions on, for the time being.

The associated booklet provides several ideas for working with the cards, with both collaborative and individual options included. The simplest is what we call the discovery mode: working in a small group, begin by drawing 2-3 question cards at random. Talk through how the questions and their keywords relate to the algorithmic system you are focusing on. Relate the keywords and the questions to one another. Once the discussion seems to be done, you can continue by drawing more cards, and either add them alongside the existing ones, or replace them to start a fresh discussion.

During the development of the cards, we tested them in workshops and group settings with people coming from different backgrounds, and they have worked very well. The REAL card deck provokes critical and open-ended conversation about ethics and values in algorithmic systems. With its new vocabulary, the deck inspires users to think differently about algorithmic systems.

The REAL card deck is openly available in figshare, under a Creative Commons license: https://doi.org/10.6084/m9.figshare.29222054

If you try out the cards, please let us know how it goes, we are eager to hear about your experience!

**

The REAL card deck was designed as part of the Reimagine ADM project by Tuukka Lehtiniemi, Maiju Tanninen, David Moats, Elisa Elhadj, Maria Eidenskog, Veera Koskinen, Dorthe Brogård Kristensen, Perle Møhl, Ajda Pretnar Žagar, Minna Ruckenstein, and Julia Velkova

Graphic design by Studio Kiss / Tuija Tarkiainen, Parisa Khojasteh