It’s possible an algorithm brought you this article. What does that mean?

Experiencing Algorithms

How Transparency Can Unlock New Experiences And Product Designs

What are algorithms?

Algorithms are everywhere and yet almost invisible from a user experience perspective. This obscurity naturally leads to distrust. Large platforms like Facebook are already responding to public concerns by allowing users to turn off algorithmic ranking. But is this enough?

Without covering too much familiar territory, algorithms are sets of rules followed in a fixed order. When carried out by a computer, these fixed rules make it possible to organize data on a massive scale. The more inputs an algorithm receives, the more it outpaces our human capacity to problem solve and make decisions. If companies want to continue to build products that rely on algorithmic decision making they will need to earn the public’s trust through greater transparency.

Dedication to more legible algorithms can transform the current apprehension into the fuel for revolutionary new experiences.

What needs to be revealed?

Transparency in algorithmic decision making is all about revealing the path of your data.

1. Training Data

Possibly the most controversial element, this is the original set of inputs used to build the algorithm. For example, an algorithm that identifies facial expressions would use a library of photographed faces to train itself. Bad training data can ruin an algorithm and jeopardize any trust in the system

2. Variable Labels

Generating labels usually requires subjective human input. This is also where natural bias can have the greatest effect. In the facial recognition example these labels might classify different facial expressions as “Happy” or “Sad”.

3. Objectives

The goals we choose for our algorithm determine how the training data and variable labels will be used.

Even simple objectives can have sweeping impacts. Our facial expression algorithm might be used when hiring, scanning profile images and only accepting applicants whose images match the “Happy” variable label. It is possible that a simple objective like “center happy faces” is responsible for the apparently discriminatory outcomes of Twitter’s photo algorithm.

The objectives tell the algorithm which outcomes to optimize for when making choices. The lack of clarity surrounding objectives fuels a large part of our apprehension.

But what happens once we have transparency?

1. Algorithms Become Trusted Middlemen

Once we know their inputs and objectives, we can trust algorithms with more information

When we know the fixed rules an algorithm is set to follow, we don’t have to speculate about foul play. Instead of worrying about an invasion of privacy or manipulation, we can feel comforted by the absence of human intervention. We can enjoy personalized experiences without revealing intimate information to another person. For more complicated tasks like emotion detection, it’s helpful to know that there is no Amazon employee judging you on the other end of your Alexa.

No single Youtube employee needs to know your interests for Youtube to serve you a specific video recommendation.

2. Algorithms Unlock New Ways For Users To Communicate

If we all know how an algorithm functions we can interact with it in ways that signal our intentions to others

Users adopt new behaviors when the algorithm’s rules are common knowledge. We can see examples of this happening on platforms like Tik Tok already. Under the rules of the algorithm, content with higher engagement performs better and is more likely to go viral. You might see comments like “What’s your favorite color?” under certain videos meant to invite responses from users who also want to boost engagement with the post. The same way inside jokes build community, comments like these can be seen as displays of trust in the platform and a bonding experience.

When the algorithm is a known part of the experience, users adopt behaviors that signal their familiarity with the rules.

3. Algorithms Let Users Enter New Situations With Confidence

Algorithmic design can make a new user’s experience just as valuable as an experienced user

New products like Lunchclub are centering their experience around algorithmic decision making. Their service helps users network by placing strangers into 45 minute video calls based on their shared a goals, profession or interests. Pairing is handled by an algorithm and from personal experience it works surprisingly well. The onboarding process becomes way more engaging once you know how your information will be used and for what purpose. Even though Lunchclub is not fully transparent with their algorithm’s features, they do present its role upfront which lets users enter their first experience with more confidence.

Letting the user know that A.I. (algorithmic decision making) will be utilized makes it easier connect with the experience of the product

What Should We Do?

How do we bring attention to something ubiquitous and invisible?

The solution lies in better user education and a change in design philosophy. User education is pretty straightforward but changing design philosophies can be tricky. Overall it requires a rethinking of how to include the user in the experience of your product. Similar to the story of how Betty Crocker made a better product by adding steps to the recipe, revealing the effect users have on their own experience can unlock new value.