Collective Algorithms for Human Interaction
Discussing and speculating algorithms for community-building in an exploratory workshop at WUD Rome 2020
_________________________________________________________________
PSA: It is a key value for us at AIxDESIGN to open-source our work and research. The forced paywalls here have led us to stop using Medium so while you can still read the article below, future writings & resources will be published on other platforms. Learn more at aixdesign.co or come hang with us on any of our other channels. Hope to see you there 👋
We often see how algorithms are being used to profile us in this age of over-personalization, but how can we shift the way we use algorithms to bring about a sense of community and the collective group? What if it can be used to make a “person” out of a group? What are some fun ways that data can be used for things like team bonding? Can we gamify algorithms?
In this 1.5 hour workshop, participants were introduced to how algorithms are used to provide product personalization and given a basic overview of how recommender systems work. We discussed how algorithms to separate individuals, then in a format of a mindmap, formulated a common understanding of “community” and “collaboration”. Furthermore, we speculated ways in which algorithms can be used for human-human interactions.
Process Framework / Workshop Methodology
The approach for this workshop is to introduce participants to two lines of thought prior to the speculative design session. The first line of thought is on algorithms for personalization — what the use cases are and how the algorithms work, more specifically recommender systems. The second line of thought is on defining community and collaboration. Then in the speculative design session, participants are given assignments to help them cross-pollinate and come up with ideas and form rapid prototypes.
Algorithms for personalization
What applications of algorithm-driven (smart) systems do you know?
The use cases for how algorithms are used for personalization vary from predictive content personalization to assistants to tracking & analytics.
Based on historical data or behaviors, algorithms can predict content that would be most relevant for users. From Spotify playlists and song recommendations to Netflix series suggestions, users dynamically receive
AI assistants can take shape in the form of voice-based or chat-based. Technology like Siri and Alexa makes it possible for each individual to have their own personal assistant. Vue.ai provides users with a personal fashion styling assistant. Albeit not (yet) as smart as human assistants, they are available at the reach of the hand.
Algorithms have also allowed for tracking & analytics on an individual scale. These are most often seen in health & fitness applications. Activity trackers like Fitbits utilize algorithms to detect what type of movements is being performed by users. Apple Watch can detect irregularities in heart rates. Individuals can access deep insights on their lifestyles and behaviors just on their phone, rather than having to get tests in labs or hospitals. Beyond the realm of healthcare, some applications utilize Natural Language Processing (NLP) to analyze a writer's writing style, providing insights and editorial suggestions.
How do these algorithms work? In this workshop, we focused on recommender systems.
Recommender Systems
Before algorithms, recommendation used to be a social act that drives relationships, whether professional or personal.
How do recommender systems predict personalized content? A basic approach is collaborative filtering. This technique uses the historical data of users to determine recommendations. Based on users with similar preferences, the system would recommend a product to another user. This is based on the assumption that users who have similar preferences in the past would also have similar future interests.
Another approach is content-based filtering, which is driven by product similarities. Features that determine product similarity could be descriptive data about the product or product image. For example, movie recommenders could use data of genre, cast, director, release year, etc.
Recommender systems can be driven by a combination of algorithms. Within collaborative filtering and content-based filtering, there are also various algorithms that can be used.
What is community and collaboration?
Together we defined these two terms, such that we have a mutual definition for these terms as we move on throughout the workshop.
What is a community?
A community is a group with a shared vision, values, interests, and/or needs that support and help each other out. Within a community, there may be different roles, yet everyone has a sense of membership and belonging, allowing for collective action.
What is collaboration?
Collaboration is teamwork and requires active participation from parties with possibly different specializations (or communities) in order to work towards shared goals. Collaboration may result in reduced workload and a higher chance of success.
How do algorithms separate individuals?
Participants were prompted to discuss this question: How do personalization algorithms separate individuals (or not)? Important and interesting points were brought up.
1. Reinforcement of societal tendencies to polarization
Recommendations based on user profiles do not share goals or needs as communities do. Individualization is promoted. This reinforces tendencies to polarization which we see in society. "We as humanities need to get in touch with different realities, but algorithms are not helping to do that." By simplifying people into "labels" certain groups are marginalized and made trivial.
2. Segmentation results in one perspective
"You cannot get out of the content to explore something new, because you are already in that cluster. What is suggested to you is already connected in some way, so you get just one perspective. The exploration part is limited because you're segmented in some way."
Algorithms put each person in their own bubbles. Not only are you in your own bubble of content, but you also do not have knowledge of other people's content. You cannot see what others see and there is no way in which we can share information on the same level.
Furthermore, algorithms continue to reinforce a single perspective. “You know more and more about your group, but you know less and less about other groups. It’s where communities go to die at the end of the day.” Minority groups are also often isolated due to this reinforcement.
Although some applications, like Spotify and TikTok, do suggest content outside your bubble of interest every now and then to gauge if users would be interested in the new content. But is this really enough?
3. Locked into certain groups & isolation of minorities
Even if you have communities and groups that you belong to, you are not that free to choose them. For example, Netflix gives top trends based on your country which you are locked in to. What happens for people with multicultural backgrounds or expats looking to also enjoy trends from their home countries? It is often difficult to choose preferred communities and if possible, users are often prompted to choose only one. There is no possibility of mixing communities or groups.
Futures Design and Speculation
For this part of the workshop, participants were divided into small groups of 3, focusing on the theme of their choice. Three exercises were given to aid participants in generating ideas and prototypes, together with two themes to provide more focus for speculation.
Themes
1. Group personalization
What would it look like if algorithms are personalized for a group identity?
Eg. a Family-driven series recommender with stimulating discussion questions.
2. Gamification for collaboration/connection
Can we gamify these algorithms? How can we use them to create a (multiplayer) game?
Eg. Guess what your friends’ recommender is recommending them.
Exercises
Exercise 1: Mapping situations — What would it look like if algorithms are personalized for a group identity?
Exercise 2: Imagining futures — What would it look like if algorithms are personalized for a group identity?
Exercise 3: Prototype — Participants then selected an idea and visualized it through a rapid prototype.
There were 3 templates available that participants could use to create rapid prototypes: a Kickstarter campaign, an application, and an event page. This rapid prototyping idea was taken from a prior speculative design workshop from Speculative Futures Rotterdam and AIxDesign.
Outputs of the session
Due to the time constraint of the workshop, the speculative design session was quite challenging for the groups to come up with ideas and finish a prototype. Nevertheless, here are some of the ideas that came out of the session.
Neighborhood Dance
What if neighbors dance to songs recommended for public space within the neighborhood to get to know each other?
Movie Chat
What if the system helps us generate a fun discussion about a movie?
Xmas in Peace
What if an algorithm facilitates creating the perfect Christmas dinner atmosphere to avoid arguments?
Reflections
This was the first time this workshop was ever given and the aim was to approach algorithmic design from a different perspective. Through questioning, discussions, and mind maps, we critically examined the effects of personalization algorithms on society. We began to explore ideas for how
Although participants found the speculative design session challenging, we began to explore ideas for how to use algorithms to promote connection and community. For future workshop sessions, more time should be provided to allow participants to wrap their minds about the exercises and accommodate brainstorming and discussion within working groups.
Further resources & Readings
On recommender systems
Take a look at these resources below to learn more about how recommender systems work, the algorithms that power them, and how companies use/test them to provide personalized services.
On social impact from technology
These resources explore how personalized technology plays a role in our society.
- What are the ethical challenges of recommender systems and their effects on the society?
- An overview of the potential and issues of personalization systems from various perspectives.
- A research initiative to keep an eye on
- What does it look like when algorithms are applied to promote community and human-human interactions? Here is a good example fromIDEO.
About AIxDesign
AIxDesign is a place to unite practitioners and evolve practices at the intersection of AI/ML and design. We are currently organizing monthly virtual events, sharing content, exploring collaborative projects, and developing fruitful partnerships.
To stay in the loop, follow us on Instagram, Linkedin, or subscribe to our monthly newsletter to capture it all in your inbox. You can now also find us at aixdesign.co.
