DANI

Our Partners 

Our solution expands upon past research at Stanford University, conducted by Professor Elizabeth Murnane. The development of the tool was conducted by the Empower Lab at Dartmouth College, whose mission is to design, build, and study the impact of interactive tools that empower people to engage with data and information in everyday life. The project was done in collaboration with ATLAS Institute at University of Colorado Boulder.

UX Research

Once we learned our design challenge and the scope of our project, I knew I needed more context and data to help us ask the right questions and further understand the context of introducing AI concepts and ethics to children.

I employed the use of both secondary and primary sources to better collect data and insights in order to deepen my understanding of the problem space.

Research Method 1: Secondary Research

A deep dive into AI literacy and early childhood education

Next, I researched how AI education has posed fundamental challenges to early childhood education.

• WHY is AI necessary and appropriate for learning in the early years?

Inclusivity: We must ensure that all children and their families, especially those from less advantaged backgrounds, have access to AI technologies. In fact, early exposure can promote self-efficacy and interest among female and minority students.

Guidance: Children are not able to understand how AI-enabled devices work without effective guidance. Only through introducing AI in an age-appropriate manner, children can acquire a healthy appreciation for its abilities and limitations and develop an appropriate relationship with it.

• WHAT is the subset of key AI ideas and concepts that can be learned by children?

AI literacy can be defined as knowing and understanding the basic functions of AI and how to use AI applications in everyday life ethically. Literature on AI education for young children shows that any framework needs to focus on the following elements:

1. Knowledge-based systems

2. Supervised machine learning

3. Generative AI

 I was then was able to brainstorm different games to address the appropriate AI concepts.

It is also suggested that learning activities can be designed to help children understand how an intelligent agent (e.g., a robot) can perceive and make sense of the problem. The key to achieve this goal is to make the intelligent agent perform a transparent process of AI reasoning, so that children can better understand AI practices.

• HOW to engage children in a meaningful experience that allows them to acquire these fundamental AI concepts?

Increasing accessibility to low age ranges: Early parent-child interactions play a fundamental role in the child’s social-emotional development as well the child’s social competence. Additionally, an older family member supervising the younger child’s learning will also help alleviate the “literacy requirement” of any interface that do not require prior programming experience and skills. .

Parents as starting point: It is imperative to also use parents as a starting point for discussions in AI ethics, a core competency of AI that entails being able to understand the ethical aspects related to using AI tools.

click to view the table

Research Method 2: Primary Research

Connecting with the Upper Valley community

Our primary research included conducting semi-structured, yet conversational interviews with the children from the Upper Valley and their parents. We partnered with Montshire Museum of Science, Howe Library, and local Upper Valley families to run our interviews and gather user insights.

Our visits to the museum and the library were an immersive experience that gave us the opportunity to become acclimated with the kids and build rapport with them. During our weekly meetings and co-creation sessions, I involved children and their guardians in the defining and brainstorming steps of the design process. Based on the insights I shared and the barriers they identified, I brainstormed potential solutions with the children and gathered themes based on their proposed ideas. We used these sessions to better understand their aspirations for the solution we were designing with them, as well as the extent to which they would want to engage with it.

I also interviewed game developers from Disney Media & Entertainment Distribution to better understand the AI gaming industry.

Click to view what our users had to say

Research Analysis

I used a range of design thinking tools to visualize the influence and impact of the range of stakeholders we were interviewing. This not only helped us outline all of the various people that would interact with our platform, but how to best focus our design efforts.

From creating a user persona representative of the target market, I found that our solution had to:

• Be a family-centered educational minigame platform

• Be designed for iPads, as most kids are comfortable with interactions like tapping, swiping, and zooming.

• Have a voice-based interface (VUI) that allows the pre-literate users to understand and interact with the games’ instructions or questions

• Have a readable-script for supervisory family members, especially, for the ethics module

• Use images and symbols over written-text to assist pre-literate users with independently identifying the appropriate features

• Promote discussions in AI ethics through story-based learning

Click to view our user persona

Prototyping & feedback

Meet DANI: An Intelligent Agent For Kids To Train 

I began the prototyping process by creating different guiding agents. It is easy to reduce AI to the image of a robot, but applications for AI are far wider reaching. I explored different characters and ended up choosing a dragon.

Untrained dragons can cause a lot of damage. Likewise, as AI systems spread further and have more influence over our lives, it’s getting far more important to make sure they’re properly trained. Bias can creep into the reasoning of AI very easily, either via datasets that are not diverse enough or through irrelevant data attached to viable data points, leading to flawed results and in some cases prejudiced or dangerous conclusions.

Information architecture

We approached our prototyping process with the intention of constantly iterating and seeking feedback from the child-parent pairs and the game developers. As stated in our solution, our research indicated a prominent need for creating a digital minigame platform that is family-centered. The platform is made up an onboarding module, two minigames (AiSpy and Tree Explorer), and an AI Ethics module. I designed both modules and one of the two minigames (Tree Explorer).

I began by categorizing and structuring the information based on what was most useful, usable, and user-friendly. Accordingly, I created a user map in order to visualize a user’s path through the platform, first through the onboarding module that introduces DANI’s story, AI literacy, and logins players, the mini-games that targets ‘Searching and Decision Trees’ as an AI concept, and the learning module which exposes families to learning about and discussing different topics in AI ethics.

I created prototype plans to help facilitate this process. As we prototyped our solutions, we gave iPads to children and allowed them to navigate throughout the platform with their parents. As they explored the platform, we asked them to self-narrate any thoughts and observed their reactions. 

Iterations of AI agents

Click to view the on-boarding module userflow

Prototype 1:

Key features: 

• Onboarding module: Users can create and delete accounts

• Tree explorer: Users choose a fruit from a list and then answer Dani’s Yes/No questions.

• AI Ethics module: Users choose from a list to topics, such as machine learning bias, unemployment…etc., and users are then taken to websites that explain the concept adequately.

Feedback:

• “This aesthetic is dull and uninteresting.”

• “It is too easy”.

• “Can I skip the ethics one? It is too boring. I want to play another game instead.”

The audio of the video is copyrighted

Design decision breakdown

Aesthetics

The children were very excited about the clip art feel of the games. This prompted me to create tens of vector illustrations with Figma, making learning feel more playful and engaging.

Usability
Interacting with VUIs requires little or no training because voice is a very natural form of communication for children and more intuitive than using a keyboard and mouse. Literacy, typing skills or tech knowledge is no longer a prerequisite for fast and effective access to information about AI.

Communication 

According to Pask’s Conversation Theory, successful teaching generally involves a two-way conversation. Using the AI Ethics module, the parent adjusts to the child's needs and ability and engages the child in a conversation, to diagnose problems and misunderstandings, by asking questions and engaging in collaborative argumentation.

Additional Deliverables

Impact

Launch

This is an ongoing project at the Empower Lab with challenges that will evolve throughout development, but I am excited to see Dani’s platform launch in the Winter of 2024.

Outcome

The lab conducted more usability testing in winter and spring of 2023 with the same child-parent pairs, and so far, Dani resulted in a 50% increase in a child’s ability to understand and explain AI concepts. All children involved in the study, when quizzed about AI concepts, were capable of correctly answering questions about AI concepts that were tested. Not only did they retain the knowledge they gained from playing the game, they also felt comfortable trying to explain concepts that they did not learn about.

Reflection

• This was my first time leading UX Research as well. As a research lead, I decided the research methods used and helped determine the direction of design based on insights. I also wrote interview protocols, designed testing sessions, conducted them with participants using both generative and evaluative research methods, and led the synthesis and analysis of collected data to reveal insights that would inform our design solution.

• This project challenged me to empathize with a group of people that I was unfamiliar with. Working with children taught me to be flexible and creative in my approach. It was almost impossible to predict what the kids were going to say or do during the interviews, so I had to roll with everything. Although working with the kids required a lot of patience and persistence, the user interviews often felt like a play area more than a testing session.

• This project also taught me the complexities of designing novel products that require a large level of commitment. It is only after we take the time to understand the thoughts and feelings of the user that we can effectively support them through these processes.

Figma Pages