Understanding College Students' Experience on Snapchat's My AI

Iowa State University | Feb - May 2024

Project Overview

The Project:

As part of my graduate course at Iowa State University, I collaborated with a teammate to gain a deeper understanding of potential facilitators and barriers to the adoption of Snapchat My AI feature by college student Snapchat users.

Our holistic goal was to gain insight into user attitudes and perceptions towards GenAI chatbot features in social media.

Summary:

  • 🎭Role: 
    Interviewer, Note-taker, Create instruments (diary study tasks, diary study log template, interview guide), Analyze data, Present results
  • 🗓️Timeline: Feb – May 2024
  • 🔎Tools Used: 
    Qualtrics, Excel, MS Teams, Dedoose
  • 📏Methods: 
    Diary Study, Interview, Thematic Analysis
  • 🧑‍🎓Target Audience: 
    GenZ University students, Snapchat users

Snapchat My AI

With Snapchat’s My AI, Snapchat users can:

  • Chat with the AI Bot using text, pictures, and stickers
  • Give a name, avatar, and a bio to the AI Bot
  • Change chat background images, colors, etc.

🚩The Problem

The social media industry is quickly embracing the GenAI trend, giving GenAI access to billions of people worldwide. The early adoption of GenAI features by social media platforms like Snapchat and Meta is generating both praise and criticism from users and the media.

While the potential benefits are promising, there are several risks, including trust, privacy, and misuse. Therefore, further research is needed to ensure that the integration of GenAI in social media is both safe and beneficial for everyone involved.

🎯The Goal

Our objective of the project was to understand user attitudes and perceptions towards My AI for GenZ college student Snapchat users in the US. Our primary research questions were:

  • How do college students perceive the facilitators and barriers to the adoption of the Snapchat My AI feature?

  • What role does trust play in their perception and acceptance of Snapchat My AI?

🛣️Methodology

Diary Study

We employed a 4-day diary study where the participants were asked to record their use of My AI once a day. Each day, they were provided a unique task as listed below:

  • Day 1: Utility-based
    Recommendations for travel
  • Day 2: Entertainment
    Play a game
  • Day 3: Utility-based
    Recommendations for shopping
  • Day 4: Social
    Interact using pictures and stickers
Additionally, participants shared their motivation for using MyAI, their perceptions towards the interaction with MyAI, and screenshots of their conversations with the chatbot.

Interview

We conducted a follow-up interview with the participants to further explore their expectations, interactions, and sustained use of Snapchat My AI.

Screenshots of participants’ interactions with MyAI during the diary study

👩🏻Participants

(used pseudonyms)

👨‍🎓Michael Anderson

  • Male
  • Undergraduate student
  • Daily Snapchat user
  • Has used Snapchat My AI before the study

👩‍🎓Priya Sharma

  • Female
  • Undergraduate student
  • Daily Snapchat user
  • Has used Snapchat My AI before the study

👩‍🎓Ayesha Khan

  • Female
  • Undergraduate student
  • Daily Snapchat user
  • Has not used Snapchat My AI before the study

👩‍🎓Mei Lin

  • Female
  • Graduate student
  • Weekly Snapchat user
  • Has used Snapchat My AI before the stud

👩‍🎓Emily Walker

  • Female
  • Graduate student
  • Daily Snapchat user
  • Has used Snapchat My AI before the study

👨‍🎓Raj Patel

  • Male
  • Graduate student
  • Daily Snapchat user
  • Has used Snapchat My AI before the study

📊Findings

Diary Study

22 out of 24 diary study entries were collected from the 6 participants. We performed descriptive statistical analysis for the quantitative data and inductive qualitative data analysis for the open-ended questions. Some of the notable findings from the diary study are:

  • The participants found My AI most useful for getting travel recommendations.
    (Mean Usefulness Score: 4/5)
  • All the participants found using My AI easy.
    (Mean Difficulty Score: 1.5/5)

 

Word cloud of open-ended question responses of the diary study

Thematic Analysis: Interview

What participants saw and heard from friends and media about My AI sparked curiosity towards My AI.

“I was just like watching Tik Toks and like people were sharing screenshots about

what they’re talking and I wanted to try it.

Emily Walker

“..sometimes people would ask MyAI to, I don’t know, create a song when it first
came out, they were trying to see what they could get out of my AI. So then
there would be like people asking it, hey, can you create a song or can you..I
don’t know..help me write this essay or something, and I thought that was
interesting, that it could do that. So I thought, oh, let me see what it can do
and I think I remember asking it. Can you write a song and they came back with
lyrics, I think, or something like that.

Mei Lei

Participants’ expectations shape their interactions with My AI and thus influence the perceived usefulness of My AI.

Preconceived Notions and Expectations

The participants had preconceived notions and expectations about My AI that stemmed from:

  • Comparison with tools like ChatGPT, Google
  • Visual and interaction design elements of My AI
  • Prior experience with Snapchat

“I think for ChatGPT is like more it’s…I don’t know..I think it’s serious like
because it doesn’t have, you know, it’s it’s technically made for academic stuff.
So I think that’s more sophisticated. I don’t know if that’s the right word, but
yeah. But, MyAI is so chill; we can use emojis with it. We can we can use
anything.

Priya Sharma

Expectation versus Experience for My AI

Each participant interacted differently with the Chatbot based on their preconceived notions and expectations of My AI. Then, they measured the responses by the chatbot against their expectations. Sometimes the chatbot was able to meet their expectations and in some cases, MyAI failed.

“So in those situations I felt like I had the wrong expectations for it. Like when I
was expecting it to act more like a human and it starts talking to me like a
company or more like a machine.

Mei Lei

Perceived Usefulness and Intention to Use

Participants formed their perceptions about the usefulness of the chatbot based on how well My AI met their expectations. These perceptions influenced their intentions to use or not use My AI. Also, users found My AI helpful for certain tasks (travel recommendations) but not useful for other tasks (shopping).

“If you’re making a recipe and you might not have a certain ingredient, I think
MyAI would be good to say “ohh like what’s a good substitute for this or like ohh
I messed up this part of my recipe..how can I fix my cooking from here” because
that’s like a more specific thing that like maybe Googling can’t help with.

Michael Anderson

The amount of effort required to use My AI effectively is closely tied to how users adapt their prompts to their goals.

“I don’t think it required much effort, just I don’t know, I would ask it maybe a second question or a third question.

Ayesha Khan

“But for the soccer cleats one..so it took like will more than like ten tries to get even get like close to it….I think it depends on my experience. It depends on
what you’re actually trying to ask out of the MyAI feature.

Raj Patel

The trustworthiness of My AI is influenced by issues related to privacy and its reliability.

Trust: Privacy Concerns

“I feel like I am a little concerned of sending it some photos of myself, but I feel
like these concerns are like unfounded because at the same time. I’ve already
made peace that like this Snapchat model knows all the photos that I’ve taken on
Snapchat in the last seven years. So why am I worried about sending it another
photo of myself?

Mei Lei

“I wouldn’t trust getting too personal with it just because I don’t know if, let’s say
Snapchat, or MyAI or something? The data would get breached.I don’t know..I
mean, it happens occasionally with other companies, so I..I don’t know, I wouldn’t
ever feel comfortable being too personal with it.

Ayesha Khan

Trust: Reliability

“I feel like I could trust it as a jumping off point. Uh, so for example, for the
itinerary activity, I would use the things that it recommends and like I would
Google those places to do a little more research on my own to see if those places
are like places I actually want to go to.
Mei Lin

“Well, after it told me the bakery was Mexican based and then said just kidding, it’s actually a French-based. I don’t know. After that it was like yeah. Now I don’t even know if I trust any of your responses.

Emily Walker

Recommendations

Policy

  • Prioritize user privacy by upholding ethical standards.

Design and Development

Improve:

  • Accuracy and relevance of conversations
  • Interactivity of conversations
  • Ad Experience