As my Masters Thesis project, I designed and developed a game research tool called Merlynne on Unity to understand the effect of avatars and their visual stereotypes on persons' prosocial behaviour in virtual environments. Merlynne asked players to give advice to others using a CBT format based off of Dr. Robert Morris's Panoply within a fantasy game environment. The thesis explores the efficacy of gamification using avatars and familiar narratives in persuading cognitive emotional support using mixed-methods. Proposal has been published at CHI 2018.
Currently in submissio to CSCW 2020, but the thesis is published on UWSpace.
As my Masters Thesis project, I designed and developed a game research tool called Merlynne on Unity to understand the effect of avatars and their visual stereotypes on persons' prosocial behaviour in virtual environments. Merlynne asked players to give advice to others using a CBT format based off of Dr. Robert Morris's Panoply within a fantasy game environment. The thesis explores the efficacy of gamification using avatars and familiar narratives in persuading cognitive emotional support using mixed-methods.
As a game developer and designer, I used the Unity engine (C#) and Visual Studio to create Merlynne and Photoshop CS5 to render the art assets. Google sheets was used to store player data from Unity.
As a researcher, I conducted a mixed-methods study using automatically logged usage data from Unity and semi-structured interviews with players.
Project Overview
Researchers have explored how online communities can be leveraged for peer support, but general disinterest and a lack of engagement have emerged as substantial barriers to their use in practice. To address this gap, I designed Merlynne using the Elaboration Likelihood Model (ELM) and games user research, a serious game that motivates individuals to become peer supporters using the Cognitive Behavioural Therapy (CBT) techniques, through play.
I conducted a mixed-methods, exploratory study to evaluate Merlynne's design and specifically studied the Proteus Effect, hypothesizing that players using a stereotypically helpful avatar would have higher usage rates and a higher change in helping attitudes scores (HAS) than players using a stereotypically unhelpful avatar.
Problem
The Proteus Effect is when an individual is influenced by the stereotypes of their avatar in virtual spaces to act differently in reality. I wished to determine if using avatars with stereotypically helpful appearances can persuade male-identifying individuals to participate in peer to peer support in a game interface.
Findings were to be interpreted as design suggestions to improve online mental health platforms engagement, an identified problem in literature and industry.
Approach
In my Msc, I wished to apply my knowledge of mental health products and learn game development.
Determining Content
Early discussions of the concept of a gamified peer to peer support interface was deemed of research interest by local psychiatric centre as well as school boards, with additional interest in the way male-identifying individuals used online peer support services. We originally wished to use real reddits posts on university subreddits -available publically- however, due to ethics limitations of our institution's research ethics board, it was important that identity be untraceable. Instead we paraphrased reddit posts to be representative of themes discussed on choice subreddits. Participants were convinced of its authenticity in interviews.
Iterative Design
Merlynne was created using iterative design, with 4 MVPs of different game concepts made and roundtable discussions by HCI experts generating feedback. Early prototypes of Merlynne then underwent playtests with graduate student colleagues using talk-alouds and interviews. Confusions in flow and causes of breaks in immersion were identified, and solutions were applied between each playtest.
For the purpose of my MSc research on the Proteus Effect, two avatar conditions (Monster and Cleric) were created based on literary research from the media arts field with feedback from peers studying language in gaming.
Research
A 2 hr in-lab study was held, with 36 male participants recruited from a graduate mailing list and posters from the University of Waterloo campus. Participants completed pre-post surveys containing HAS scales before and after a 30-40 minute play session using a random avatar condition, followed by a semi-structured interview.
Peer support responses collected in-game from participants were assessed for adherence to CBT format by three researchers (with high inter-rater reliability). Data-driven thematic analysis using open and axial coding was used to interpret interviews, supported by non-parametric tests (Mann Whitney U) of usage data including aformentioned responses coded for CBT format adherence.
Results
Full paper is currently in submission to CSCW 2020. - check back June 2020
Impact
The work on Merlynne has:
- Extended previous work on the Proteus Effect and crowdsourcing CBT
- Developed an artifact for HCI research
- Identified areas of future research in avatars
Reflection
My thesis was successfully defended, and the Merlynne tool was packaged to be accessible to future researchers looking into peer support through RPG interfaces.
What I would have done differently:
- Selective sampling of persons knowledgeable of RPG mediums to test the Proteus Effect specifically. The Proteus Effect requires the user to recognize stereotypes associated with avatars. Although tested for, many participants revealed they were not familiar with fantasy rpg archetypes in interviews. This change will reduce the generalizability of study results, yet as our purpose was exploratory study focusing on the presence of the Proteus Effect, selective sampling will provide a more accurate measurement.
- Modded a pre-existing game to reduce barriers for immersion due to low fidelity of aesthetics and design.
- Involved the real concerns of individuals, and asked them to validate the effectiveness of the peer support received. Although researchers rated the reponses, we could not measure for true helpfulness given they were not the original posters.
- Recorded gameplay footage either through video or keylogging to understand the participants' gameplay choices. Where the player was clicking, travelling with their avatar, and how they structured their responses was not documented in our study, but the qualitative data could hold insight into the usage data. For example, if a player spent longer than usual on a level, were they exploring the map, rethinking their responses, or just idle? Eye-tracking would also show player behaviour, especially in measuring the amount of attention they gave to the avatar but that would have higher set-up costs and risk inconsistencies.
Specific to technical development, there were many Unity techniques I later learned which would have benefited Merlynne such as linking all text objects to a database (e.g. Firebase, or a spreadsheet) for easier content management, using tags to classify NPCs, and using Scriptable Objects to manage the status of interactive objects.
I would have also stored the incoming data natively to an excel spreadsheet to reduce the risk of losing Google conenctivity or the status of my account if it were to be used by future researchers.
Next Steps and Questions
The following questions were identified:
-
How familiar should a person be with an archetype for the Proteus Effect to occur? If the Proteus Effect occurred and suggestions were made to designers about which types of avatars to use, recent and classic media trends would need to be considered. If a current trend was to rewrite a traditional archetype (ie. a helpful monster or a damaging cleric), how long would this trend need to exist for to be relevant to the Proteus Effect? Alternatively, if designers were to build a character behind an avatar, how well would the player need to be familiar with the character?
- What is appropriate feedback for peer support? Participants anticipated more motivation to help if their responses were validated to be helpful, either in-game or otherwise. Designing appropriate feedback that encourages peer participation without encouraging harmful participation would be an area of future research
- Are there gender differences? Given our participant pool of male-identifying individuals, other genders may yield different results. The relationship between genders and gendered avatars make be of interest.
In final reflection, I believe the philosophical question of whether peer support should be encouraged with gamification needs to be revisited, as it lessens the need for reflection of impact for continual help. Such a question begs attention from mental health experts.