Ghost Hunters

A Persuasive Intervention in the form of a social detection game targeted at younger social media-active adults to develop critical thinking skills and improve their ability to recognize fake news.

Ghost Asset from Game
Overview
Project Goals
In fall of 2020 I worked with a small group of students at Carnegie Mellon University to produce a prototype mobile game that used persuasive techniques to increase users’ critical thinking skills when detecting fake news. This project was completed over a semester for the Persuasive Design course at CMU.
Group Composition
At the beginning of the semester we were assigned small groups based on common interest in intervention topics. I joined the group focused on improving online behaviors along with fellow MHCI students Meo Zhang, Raajat Gupta, Zach Van Cleef, Aziz Ghadiali, and Cydney Vicentia. We all brought different experiences to the table and chose different project roles. I focused mainly on research and writing, as well as producing our final video presentation.
CMU HCII Logo
Exploring the Problem Space
Background Research
Screenshot of the user testing presentation grading articles based on reliability metrics
To accurately represent the issue of Fake News and understand what aspects of fake news were most relevant for our intervention, the group began with several weeks of intensive background research. Having experience with psych literature I began the lit review, and with contributions from other members we compiled a solid basis of peer-reviewed information. Additionally In assisted in conducting several expert interviews with researchers in the field. Other group members found active Facebook users and interviewed them along with administering a small exercise to adjudicate their ability to spot fake news.
Findings
Abstraction of a Feedback Loop
Overall our findings indicated that many users wind up sharing and viewing questionable articles thanks to a phenomenon known as a feedback loop. These cycles of information see small clusters of people, who form like-minded in-groups on social media, sharing and posting information that reaffirms their beliefs. The more congruous an article is with the group’s beliefs, the more likely it is they will accept it without question, even if it is blatantly falsified. We saw this in user testing when our younger, liberal-leaning users decided certain articles were perfectly valid to be shared without even reading them or checking the source of the article. This tipped us off that a good target population should be these younger socially conscious users who understand that fake news is a problem but lack the skills and resources to detect it.
Intervention Targets

We decided that our intervention would target younger, socially-conscious users who had a pre-existing understanding that fake news was a problem but did not possess the critical thinking skills to differentiate it on their own social media feeds. From our user testing and literature review we determined that many young social media users fit within this category, and would be more feasible to persuade than older online populations who are less concerned with the fake news phenomenon.

From our background research it was clear that many avenues of intervention could be taken to help reduce the spread of fake news. While feedback loops were a large issue, convincing people to expand or leave their social circles was a huge and ultimately unrealistic challenge. Instead, we wanted to focus on training users to more accurately spot the signs of a fake news article, which our user testing found to be a deficiency for many people.

After collaborating on an extensive affinity diagram our group decided on the idea of a social detection game with mechanics that would subtly teach users critical thinking skills in reading news articles as they played. Our group liked this idea due to the high crossover between our target user (young, highly online people) and the game genre (including popular games such as Among Us).
Design Iterations
Brainstorming Rapid Prototypes
At this stage of the project our group had a good idea of the target user and persuasive goals of our intervention, but we needed a solid prototype to test with. We began by splitting off and creating several quick ideas in tandem. From these early iterations we decided on a game where players would assume the role of ghost hunters searching an abandoned house for rituals to exercise a demon, with one player secretly playing as the demon and attempting to trick the others into failing the exorcism.
Home screen of Prototype
Gameplay Design
Evidence screen showing news article and user options
Our intervention revolves around a gameplay loop of finding evidence, determining if it is valid, and bringing it to the group discussion phase where players argue for the validity of their piece and vote on one to exclude from the ritual. When designing the prototype I worked with one other group member to create all of the written content. This was a crucial element as we needed the articles to be entertaining for players to read while still containing the relevant aspects of news articles that we wanted to capture (source, headline, image and text content). I worked to balance the evidence so that some were written more like trustworthy news articles and others borrowed fake news red flags defined on Snopes.com and from literature we reviewed. These red flags could include missing information such as an author with few credentials, a ‘clickbaity’ headline with only a picture or short description, or come from an unknown site.
Persuasive Elements
Our group worked to ensure that the intervention utilized many research-supported methods for persuasion so that users would be likely to retain the critical thinking skills they learned during gameplay. This embedded design included obfuscation via a fictional world and competition-based gameplay, as well as activation of system 2 ‘conscious’ thinking when viewing online news.
Pre-round screen telling user their role in gameplay
User Testing
reference image of axe
Low Fidelity
model of axe
Mid Fidelity
model of axe in room
High Fidelity
Our prototype went through 3 rounds of user testing. In the first round we tested a variety of game concepts before narrowing down to one that incorporated the best of everything based on user feedback. We found that many users were confused by the rules we provided, making clarity a priority in future iterations. We also found that people were fatigued when reading too many articles during gameplay, so we put a focus on intermixing our main intervention with other game mechanics to make the users more comfortable when playing. In the second iteration, participants positively reacted to the social dynamic of the game, which we simulated by having group members play along as other players during the session. We wanted players to work together and vote on a piece of evidence that seemed the most suspicious, therefore getting the user practice at singling out unreliable sources of information. These mechanics were still being fleshed out and many users reported being confused by this process at this stage. Our final prototype was still operated via wizard-of-oz mechanics by group members behind the scenes, but were able to test pairs of participants in single sessions, showing us how true group dynamics would play out within the game. Our findings were overwhelmingly positive, with players working together to discuss the merits of their collected evidence and even using tactics that were directly applicable to the real world, such as researching the source of their evidence to find if it was reliable.
Wizard of Oz Prototpying Setup in Figma
Final Presentation
Our group presented the results of our user testing and the mechanics of our intervention to our class in December of 2020. I produced a 30-minute long video that detailed our creative process and showcased our high-fidelity prototype.

Link to the Figma Prototpye
Takeaways
Background used for virtual presentation of the intervention

This project required intense planning and coordination between the team members to properly test our prototype with the users. Several team members had to work behind the scenes to line up elements of the different screens based on real-time choices made by multiple users as they played through the prototype, while others led them through and some acted as confederates. It was a challenge but I learned a great deal about coordinating such efforts and feel comfortable with user testing in such high-pressure situations.

We found that the more our users could see through to “the point” of the intervention, the more they would find reading the articles to be dull or tedious. From these initial user test findings our group prioritized obfuscation techniques and other methods of distancing the face value of the experience from the persuasive intent, so that users felt more like they were playing a game and less like they were learning a lesson.

The amount of work hours our group put in for this project was well beyond the scope of the assignment, but due to our passion for the subject matter and good dynamic we were able to make everything happen before the deadline, including the production of over 10 prototype gameflows and three rounds of user testing in three weeks. I learned a lot from my peers about effective ways to break up workload, letting members take over sections that suited their expertise and knowing when to step back if someone else has a good idea.
Created by Brady Baldwin, 2020