MIT STEP Lab and Media Lab partnered with Boston College on a one-year EAGER project to investigate whether and how students are able to learn key AI concepts and become more interested in AI and related careers and to build field-advancing knowledge about appropriate measurements and instruments to assess middle school students’ concept knowledge, awareness of AI and perceptions about AI, and career orientation. The Developing AI Literacy (or DAILy) project specifically addresses middle school students (ages 11-13) with no prior experience with AI. Youth learn the key AI concept, investigate potential for bias in the datasets and algorithms (and potential mitigation strategies), recognize the societal and ethical impacts of biased AI systems, and connect AI to their daily lives and future selves. They are also engaged in “AI and my future” career exploration activities to become aware of AI related careers, recognize their own strengths and interests for future jobs, and realize the importance of technical skills development and the ongoing nature of change and adaptation in today’s job world. Many of the AI learning activities produced through the project are not dependent on the availability of computers, contributing to multiple pathways for broadening access to and engagement in AI learning experiences for underserved students who do not have consistent access to Internet services.
Discussion from the 2021 STEM For All Video Showcase (17 posts)
Irene Lee
Co-Presenter Director
May 11, 2021 | 08:53 a.m.
Hello and thank you for taking the time to watch our video! We'd love to hear your comments and questions about any aspects of this project. In our next phase, we have started offering Teacher Professional Development on the curriculum and would love to hear where you think you would integrate AI into middle school classes.
Ateng' Ogwel
Researcher
May 18, 2021 | 02:22 p.m.
Helen and Irene,
Thanks for sharing your fascinating project that introduces AI in middle school. I am also impressed by your focus on ethical issues and sharing results showing areas where significant outcomes were noted and areas which require further insights on whether they need to be approached differently. I'll be following further from the sites provided.
Michael Chang
Facilitator Postdoctoral Research
May 11, 2021 | 12:18 p.m.
Thank you for sharing this project! I appreciated the focus on addressing bias as a first class topic in AI literacy alongside more complex technical topics like GANs and neural networks. I am curious how the curriculum frames conversations about what students can do today in terms of imagining and building their own AI systems. As mentioned in the video, services that allow students to build their own AI systems are fairly inaccessible or limited in what they can accomplish. How do instructors and students navigate the question of how most people lack the sheer resources (if not the expertise/literacy) to build complex AI systems, but are nevertheless exposed to these AI systems everyday in, for instance, their use of social media?
Irene Lee
Co-Presenter Director
May 11, 2021 | 02:15 p.m.
Thanks for the questions, Michael. To answer your first question, I would say that this project is focused on AI Literacy for 10-13 year olds rather than AI development and conversations on what students can do to build their own systems did not come up organically. Students learned the structure and functioning of different AI models (decision trees, neural networks, and GANs) through game play and participatory simulations and they used Teachable Machine to build ML models, but they did not create models from scratch.
In your second question (How do instructors and students navigate the question of how most people lack the sheer resources (if not the expertise/literacy) to build complex AI systems, but are nevertheless exposed to these AI systems everyday in, for instance, their use of social media?), I think you are talking about the power imbalance in who has the datasets and resources to build the complex models. We didn't touch on that explicitly in the curriculum though we have had internal discussions of how to distribute that "power."
Andres Colubri
Facilitator Assistant Professor
May 11, 2021 | 12:18 p.m.
Hi Irene, thanks for carrying out this important research on the educational outcomes of AI curriculum. Was it mostly through presentation materials, or did you include some hands-on experiences using existing ML sandbox environments that students could use to experiment with ML algorithms. Regarding your conclusions, why do you think that gains in NN and GANs were limited? Deep learning could be seen as one of the more exciting topics currently (with applications such as text generation, artistic style transfer, deep fakes, etc), maybe the underlying algorithms are too complex to be sufficiently engaging to the students?
Irene Lee
Co-Presenter Director
May 11, 2021 | 01:58 p.m.
Great questions. We utilized hands-on experiences, participatory simulations, and testing out AI tools activities for each key concept. Typically, we introduced a concept, then had an experiential component, an ethics component, and a future/careers component. We hypothesize the learning gains on NN were limited for several reasons: 1) the exposure time was low (45 mins in a 30 hour curriculum); 2) the terminology was difficult for students and not repeated enough, and 3) conceptually, the processes are difficult to learn in a single exposure. Furthermore GANs learning gains were limited because they relied on NN understanding.
Andres Colubri
Facilitator Assistant Professor
May 11, 2021 | 12:19 p.m.
BTW, just a quick comment the aieducation.mit.edu does no longer work, is this one the correct page: https://raise.mit.edu/daily/ ?
Irene Lee
Co-Presenter Director
May 11, 2021 | 01:51 p.m.
Thanks for the heads up. The RAISE site went up just the other day and the aieducation.mit.edu/daily should redirect there... I'll put in a bug report. Thanks!
Hi Helen and colleagues! It's great to learn about the exciting work you are doing with AI and literacy. Your video is thought provoking. We share your interest in making science more accessible and understandable. See our work exploring the ways teachers use data while teaching for social justice in science--including helping students interpret scientific data. Enjoy, Marcia
Helen Zhang
Lead Presenter Senior Research Associate
May 13, 2021 | 11:16 a.m.
Hi Marcia, thank you for watching our video and sharing your work! Our next phase of this project will focus on bringing our curriculum into classrooms. We'll explore how to support teachers to make AI accessible and relevant to students. We'll watch your project.
Jeremy Roschelle
Facilitator Executive Director, Learning Sciences
May 12, 2021 | 07:52 p.m.
I agree with the previous comments that its great to see how you intertwined learning about AI with learning about its ethics. And its so cool that you were able to do this all during remote learning. So what's next for your team? I always love when a first project doesn't give us just an answer (e.g. students can learn this) but also a better question. What new question do you now want to ask?
Irene Lee
Co-Presenter Director
May 12, 2021 | 08:52 p.m.
Hello, Jeremy.
Thanks for your question. We are embarking on the next stage of our project in which we are focusing on developing and studying a teacher professional development program aimed at supporting middle school teachers to offer the curriculum to middle school students during the regular school day. We are investigating
the extent to which middle school students can understand AI concepts and gaininterest in AI careers and future endeavors (a continuation of the previous study with revision and refinement of the curriculum),
productive approaches to prepare teachers to support students’ learning of AI concepts and skills, and to expand youths’ notions of future selves, and
the effectiveness of the DAILy curriculum when implemented by educators other than the curriculum developers.
1
Discussion is closed. Upvoting is no longer available
Found helpful:
Jeremy Roschelle
Jeremy Roschelle
Facilitator Executive Director, Learning Sciences
May 13, 2021 | 01:08 p.m.
Seems like it will be challenging to make or find assessments of what students learn. I'd love to get updates from your team on how you make progress on the assessment issue. Good luck in your next stage!
I echo Dr Roschelle's hope to gain updates on your assessments. This is an important contribution that I think will be of great interest and benefit to many. Thank you
Irene Lee
Co-Presenter Director
May 13, 2021 | 03:21 p.m.
We have developed and used an AI concept inventory to assess student learning. Recently we conducted an IRT analysis. Revisions will follow. The analysis was submitted for publication - I'll let you know if/when it can be shared.
We would not only be interested in assessment materials, but wonder whether the content materials could be adapted for our collaborative network of students internationally as a means to break into AI more fully as a topic of self-directed project activities that kids in different countries manage together. If you have any thoughts, they would be welcome. Thank you.
Irene Lee
Director
Hello and thank you for taking the time to watch our video! We'd love to hear your comments and questions about any aspects of this project. In our next phase, we have started offering Teacher Professional Development on the curriculum and would love to hear where you think you would integrate AI into middle school classes.
Ateng' Ogwel
Helen and Irene,
Thanks for sharing your fascinating project that introduces AI in middle school. I am also impressed by your focus on ethical issues and sharing results showing areas where significant outcomes were noted and areas which require further insights on whether they need to be approached differently. I'll be following further from the sites provided.
Michael Chang
Postdoctoral Research
Thank you for sharing this project! I appreciated the focus on addressing bias as a first class topic in AI literacy alongside more complex technical topics like GANs and neural networks. I am curious how the curriculum frames conversations about what students can do today in terms of imagining and building their own AI systems. As mentioned in the video, services that allow students to build their own AI systems are fairly inaccessible or limited in what they can accomplish. How do instructors and students navigate the question of how most people lack the sheer resources (if not the expertise/literacy) to build complex AI systems, but are nevertheless exposed to these AI systems everyday in, for instance, their use of social media?
Irene Lee
Director
Thanks for the questions, Michael. To answer your first question, I would say that this project is focused on AI Literacy for 10-13 year olds rather than AI development and conversations on what students can do to build their own systems did not come up organically. Students learned the structure and functioning of different AI models (decision trees, neural networks, and GANs) through game play and participatory simulations and they used Teachable Machine to build ML models, but they did not create models from scratch.
In your second question (How do instructors and students navigate the question of how most people lack the sheer resources (if not the expertise/literacy) to build complex AI systems, but are nevertheless exposed to these AI systems everyday in, for instance, their use of social media?), I think you are talking about the power imbalance in who has the datasets and resources to build the complex models. We didn't touch on that explicitly in the curriculum though we have had internal discussions of how to distribute that "power."
Andres Colubri
Assistant Professor
Hi Irene, thanks for carrying out this important research on the educational outcomes of AI curriculum. Was it mostly through presentation materials, or did you include some hands-on experiences using existing ML sandbox environments that students could use to experiment with ML algorithms. Regarding your conclusions, why do you think that gains in NN and GANs were limited? Deep learning could be seen as one of the more exciting topics currently (with applications such as text generation, artistic style transfer, deep fakes, etc), maybe the underlying algorithms are too complex to be sufficiently engaging to the students?
Irene Lee
Director
Great questions.
We utilized hands-on experiences, participatory simulations, and testing out AI tools activities for each key concept. Typically, we introduced a concept, then had an experiential component, an ethics component, and a future/careers component.
We hypothesize the learning gains on NN were limited for several reasons: 1) the exposure time was low (45 mins in a 30 hour curriculum); 2) the terminology was difficult for students and not repeated enough, and 3) conceptually, the processes are difficult to learn in a single exposure. Furthermore GANs learning gains were limited because they relied on NN understanding.
Andres Colubri
Assistant Professor
BTW, just a quick comment the aieducation.mit.edu does no longer work, is this one the correct page: https://raise.mit.edu/daily/ ?
Irene Lee
Director
Thanks for the heads up. The RAISE site went up just the other day and the aieducation.mit.edu/daily should redirect there... I'll put in a bug report. Thanks!
Marcia Linn
Hi Helen and colleagues! It's great to learn about the exciting work you are doing with AI and literacy. Your video is thought provoking. We share your interest in making science more accessible and understandable. See our work exploring the ways teachers use data while teaching for social justice in science--including helping students interpret scientific data. Enjoy, Marcia
Helen Zhang
Senior Research Associate
Hi Marcia, thank you for watching our video and sharing your work! Our next phase of this project will focus on bringing our curriculum into classrooms. We'll explore how to support teachers to make AI accessible and relevant to students. We'll watch your project.
Jeremy Roschelle
Executive Director, Learning Sciences
I agree with the previous comments that its great to see how you intertwined learning about AI with learning about its ethics. And its so cool that you were able to do this all during remote learning. So what's next for your team? I always love when a first project doesn't give us just an answer (e.g. students can learn this) but also a better question. What new question do you now want to ask?
Irene Lee
Director
Hello, Jeremy.
Thanks for your question. We are embarking on the next stage of our project in which we are focusing on developing and studying a teacher professional development program aimed at supporting middle school teachers to offer the curriculum to middle school students during the regular school day. We are investigating
Jeremy Roschelle
Jeremy Roschelle
Executive Director, Learning Sciences
Seems like it will be challenging to make or find assessments of what students learn. I'd love to get updates from your team on how you make progress on the assessment issue. Good luck in your next stage!
Eric Hamilton
Irene and team,
I echo Dr Roschelle's hope to gain updates on your assessments. This is an important contribution that I think will be of great interest and benefit to many. Thank you
Irene Lee
Director
We have developed and used an AI concept inventory to assess student learning. Recently we conducted an IRT analysis. Revisions will follow. The analysis was submitted for publication - I'll let you know if/when it can be shared.
Eric Hamilton
We would not only be interested in assessment materials, but wonder whether the content materials could be adapted for our collaborative network of students internationally as a means to break into AI more fully as a topic of self-directed project activities that kids in different countries manage together. If you have any thoughts, they would be welcome. Thank you.
Deborah Seehorn
Great project! I really liked the focus on ethics and career exploration, both of which are important for middle level students.
Further posting is closed as the event has ended.