This project used a more subdivided method to give users more control and autonomy instead of using rude and straightforward algorithms to determine users as a chain of data in the social media.

Ovreview

A.I. algorithm is embedded in social media. The entire algorithm logic and the social media interaction experience are very efficient and guiding, which makes people more aggressive on social media and does not inspire meaningful discussions. In response to this phenomenon, I explored different possibilities to make users explore an app by themselves with more openness and autonomy.

My Role

I designed a new social media application. The algorithm logic and interactive experience in this application are all different from the current. These new features give users a chance to develop a relationship with the app through the logic and interaction of the algorithm and without defining their using behaviors as data.

Design Outcomes

Research

Secondary Research

In order to better understand how algorithms now work in social media, I conducted research on different social media which is now widely used.

I found algorithms embedded in social media platforms understand users as data, not humans. it only uses data to calculate user behavior.

User Research
In order to have a deeper understanding of the algorithm ration and influence, I did a questionnaire📄. My target audience is 18-65 year old people who would like to spend time on social media.

These algorithms have guided our emotions, public opinion, extreme ideas, human behaviors. At the same time, the algorithm also limits the content browsed, lacking diversity.

How can I re-create a new system to make users motivate more meaningful discussion and have more chance to view broader contents?

Ideation

Persona

Based on the findings, I generated 2 personas based on their frustrations and needs during using experience about algorithm in social media. I also mapped out the user journey to find their needs and motivations.

User Journey Map

Design Highlight

Recommendation: 50% user preference + 50% randomness.Give more precise result by reading the level of preference enable to recommend fragmentized and decentralized content

Create a grid feature which can be flipped to different rows.Change the layout of contents displayed to show how much users like the content based on the gradient changing interaction.

Follow concise design principle let user comfortable. Remove overwhelming and over-efficient feature to build a opened relationship and autonomy with users.

Prototype

Design Process

In order to better understand how the algorithm works on weibo, I did some behavior experiments and sketched a diagram illustrating the relationship between algorithm and contents. The experiment concludes the result data.

Goal:
How Weibo recommend the discovery contents to users

Method:
Take two same prefrence accounts viewing two different topics respectively, focusing on the viewing topics and browsing behaviors of them. Observeing If the recommended contents are ralative with the same topic browsed.

Conclusion:
After browsing about two hours, the discovery page keep recommending the same topic contents related with the one I browsed.

Insight:
As more and more relative contents appears, users may be glad to discuss under them. However, users are thinking about browsing more other contents. Actually, except searching and following, users don’t know what they’re interested in.

Idea:
Redesign a mechanism to allow users to actively explore or passively accept new recommendations. There are content that the user hasn't viewed.

Goal:
How Weibo manage the arrangement of comments under blog posted

Method:
Take two same prefrence accounts and replying under the same topic with different sides, observing if the comments arrangement are random or based on users’ preferred sides.

Conclusion:
The comments arrangement are not based on preferred sides. It depends on the number of replying and liked. Users will be recommended more about the preferred side of topic.

Insight:
As user replied more to one side, the discovery page recommend more about this side of topic. This makes users think if everyone think as i thought? Sometimes users will be influenced by the opinion with a high number of likes, although it is not correct.

Idea:
Create a spam vocabulary screening mechanism. Present all the points of view, few opinions at the top are from different sides which is allowing users to browse to more ideas.

Diagram

According to my experiment, I have summarized some insights, so I recreated the algorithm and sketched the diagram below.

Information Architecture

Visual System

Low Fidelity & Iteration

Design

Future Development

I rethought the concept and communication method of my project. I think that in future development, algorithms will definitely occupy an important part. Indeed, algorithms can never replace human thinking. We should learn to separate from the over intelligent and efficient life and stick to our thinking.

I will work harder to explore how to balance the relationship between people and artificial intelligence, making A.I. exists in our lives but without controlling our lives.