Latest / AI Advocates / Educational Research Made Easier with Eddie

Transcript

Educational Research Made Easier with Eddie · AI Advocates

• 0:00
  Music: Reclaim your time time.
• 0:09
  Lisa Dieker: Welcome to AI Advocates. I'm Lisa Dieker.
• 0:11
  Maggie Mosher: And I'm Maggie Mosher, and today we're here with David and David, I'd love for you to introduce yourself, but I first want to say thank you to Tiffanie Zaugg, who's connected us when you talk with educators like Tiffanie, who are deep into the AI integrated space, there's a lot of common talk that goes back and forth, but what I loved that Tiffanie shared with us is your background is fascinating. You really bridge two worlds. You're an education researcher and an ed tech founder, and I can't wait to talk about your tool, Eddie, but tell us a little bit more about yourself.
• 0:41
  David Blumenthal: Thank you Lisa. Thank you Maggie. And thanks to Tiffanie, the I guess the fourth partner here today, for connecting us. Happy to share my background. So for the last 13 and a half years, up until last March, I've been working at a company called American Institutes for Research, which is a behavioral social science firm that conducted conducts. They still exist Research and Evaluation in education, and it really spans the whole gamut special education, early childhood education, you know, multi-tiered systems of support, all kinds of things. My particular background is in secondary and post secondary transitions, and so things like multi-tiered systems of support, and in particular, early warning systems in education, using the ABCs. You might have heard about these from people like Bob Balfanz at Johns Hopkins University, and using those early indicators to identify students that are showing early signs that they may be off track for graduation. And I bring that up because when I had a bit of a change in my life about a year ago, about 13 months ago, I started thinking about what I wanted to do next. There's a lot of disruptions in education, research and evaluation due to the cuts to the US Department of Education, and that affected me.
• 1:53
  David Blumenthal: And I had started talking with a friend of mine before all of that, like in in February or so about, you know, there's this thing called artificial intelligence. And I had been a bit of a skeptic, you know, I think a lot of academics, I'm sure you two, have had some skepticism about academic dishonesty or doing things that may be a foul of, you know, not really our authentic work. And so when I was talking with my friend Matt, who is now my co founder of our company, Hunter's Point Research Technologies, I was asking him, like, what is this and how does it work? What does it do? It seems like this black box that, like, stuff goes in and then things come out, and then what's what's happening?
• 2:36
  David Blumenthal: How does the sausage being made?
• 2:36
  David Blumenthal: And Matt is a friend of mine that I knew just because we happen to be both living in New Haven, Connecticut, at the time, while our partners were getting postdocs, and while he has a background in computational neuroscience. So somebody that knows a thing or two about machine learning and how to build these systems, and me, didn't knowing a thing or two about education research and disseminating research and getting research out there in the hands of educators. I started asking questions about, you know, how do large language models work, and what do they do, and what's, what's the whole process? And that's kind of where the idea for Eddie all began. So it was, it was at a basketball game at Madison Square Garden in New York City, where I lived, in February of 2025. By March, I had been let go from my employer, and so I started thinking through, what am I going to do next? And that's where, where Eddie came about. So Eddie is a chatbot, essentially, but we are building out a full platform right now as we're recording this to make it a more full feature experience that essentially translates research into clear and concise recommendations based on learning science. So we built a curated database of over 400,000 peer reviewed journal articles what constitutes learning science and school improvement.
• 3:56
  David Blumenthal: The research basis behind that.
• 4:02
  David Blumenthal: Because one thing that I found in talking with educators throughout my career, and also as I've been working on this, this idea for Eddie, is that there's, there's an issue of getting access to research. I know what teacher goes home on a Tuesday night and they've got a kid that's really struggling with literacy, and they don't know what to do, and I know I'll read an academic journal like they don't do nobody does that.
• 4:26
  David Blumenthal: You two, Lisa and Maggie, you might do that, but it's not something that most teachers are going to do, necessarily. And so there's a time crunch, because most teachers are just overwhelmed with the amount of work that they have, working upwards of 53, hours per week.
• 4:37
  David Blumenthal: But there's also an accessibility issue, paywalls, knowing where to go to find high quality, rigorous academic research and understanding how to access it, and then that issue of reading through what could be very dense technical language. It's written for researchers. It's written for for and by people like us, not necessarily for and by educators, and so Eddie is designed to really get around that challenge and that roadblock. So what we've done is we've built a tool that answers questions that educators may have, and I think a lot of your listeners may have questions like, how do I build in accommodations that are really beneficial for students that are facing a particular disability, or what can I do to scaffold this learning? Or how can I structure a small group lesson?
• 5:24
  David Blumenthal: Eddie is really designed to be a thought partner and then do that generative AI. I'll use the term magic, although, like one of our competitors, has kind of co opted that term. But, you know, do the magic of generating instructional materials, of lesson plans, worksheets, assessments, all that stuff that educators need to actually deliver instruction in the classroom. So that's where the idea from, Eddie was born. And what it does is it helps answer questions that educators may have late at night or on the weekends as they're getting ready for class the next day, or even in the middle of a planning period, and then creating documents and really saving time.
• 6:00
  Maggie Mosher: Awesome. And what I want to make sure our listeners know is that know is it's E-D-D-I-E-A-I.net where you can get Eddie, because there's also another Eddie AI, and that can be confusing, so make sure it's E-D-D-I-E-A-I.net and what I did love about what you just said is that you essentially you knew the problem. From being an educator, you knew the problem and you wanted to fix for us educators. And what we love about that is love about that is our podcast tries to stay free for educators. And there's a part of your platform that is free, which we love, and then there's a part that costs $5 a month, which is also affordable.
• 6:31
  Maggie Mosher: So tell us the difference between the free platform, what is the free platform provide for educators?
• 6:39
  David Blumenthal: Yeah, absolutely. So right now it's marketed directly to educators.
• 6:41
  David Blumenthal: So if you're a teacher, if you are an administrator, we have instructional coaches that are users. We have people that are they just want to write a really quick literature review. That's where the free version is a great thought partner to quickly answer, what does the evidence say? You know, you could ask it in plain language. You don't need to know, like boolean terms, you don't need to understand, like where you can go, and all the different like drop down menus and Academic Search Premier. You could just ask it in a very clear, concise way. This is the challenge I'm facing. What can I do? And it'll provide answers and cite its sources, and so you can see the articles that Eddie is reading, and it'll share those full text citations that used in order to generate the response. So there's some transparency there around where it's getting its answer. So we're the only AI tool that is citing its sources all along the way, and it's all coming from high quality, rigorous research. We're like Tiffanie Zaugg, who built her own AI tool. We were very intentional about the database and the literature that we're using, so that we're not perpetuating disproven practices. There's a concern around ineffective approaches, and we want to make sure that we're not helping spread that information, which a lot of other LLMs may do. There is a paid version. We're actually launching an entirely new version this June called Eddie Pro, and that version does a lot of the time savings, so it does all of the thought partnership and answering questions about education research or what to do in in the classroom to improve outcomes, and it'll create documents. It will create lesson plans, rubrics, worksheets, assessments. Allow you to share those with your colleagues. If you're a teacher and you want other people in your department, or you're part of a professional learning community, like you could share those out and say, like, here's the thing that I created. The goal is to keep the teacher as the human in the loop. They're the ones who are really doing the work. But Eddie is that thought partner, and is drafting the first draft and then giving it to the educator to then refine it. So a process that may have taken hours or days even suddenly can happen in minutes, and then you can go about your day and do all the other amazing things that educators do.
• 8:59
  Maggie Mosher: And I have one more question. I'm gonna lead it to Lisa, because she's got some great ones for you. Well, you did say human in the loop, which Lisa and I love. We want to make sure that that's imperative, that human in the loop. We've been on your platform, Eddie and we liked the design we love that you're trying to fill that researched classroom gap. But for our listeners who are in the trenches every day, why should they go here instead of just a generic AI model? You said a little bit about it's better evidence based. But what is the model behind your model is, who, what's feeding your model?
• 9:28
  David Blumenthal: Yeah, absolutely. So our model is agentic, and so it's actually multiple different large language models that have been bundled together. This is where my co founder, Matt, could do a better job of explaining how all of this works. But essentially what happens is, when a prompt comes in, Eddie will then go through the database that we set up. So we've closed off where Eddie is going to look for answers. It's not going to go to Reddit, it's not going to go to Tiktok or social media to find answers. It's not going to look at teacher blogs. Those are all valuable in their own space, but we think that put. Our bet in the hands of, what does rigorous research say works in education is a great starting point. And so then it's going to find the answers there, and then it'll reformat it essentially and say, here's a summary of the answer that you may have. So example of this, you know, I was talking with with one of our early users recently, and you know, they had an incident where, you know, one of the kids got into a fight, and it was a kindergarten student, and they got into an incident at recess, and they needed to draft parent letter to the parents, let them know, and then create a a behavior plan.
• 10:33
  David Blumenthal: So what Eddie will do in this, this circumstance, actually draft the the MVP version, the or the pro version, which is, which will be coming on June.
• 10:40
  David Blumenthal: We'll actually draft the communication and say, here's a here's an email to template that you can use, and then here is a behavior plan that you could attach to the message. And within the plan itself, there's a citation to research that was conducted on behavior planning and what which kind of models of behavior planning actually lead to improvements in student behavior. So it's not basing it off of, you know, intuition so much we think professional judgment is a really critical part and why we are so insistent on building a teacher tool that keeps the human in the loop. But we want to make sure that we're arming teachers with the most up to date evidence and research and giving them tools that that show how they think. So that's why we built in the citations. I don't think you're going to find those kind of citations back to rigorous research and any other tool that's out there.
• 11:29
  Lisa Dieker: So two questions, one is, I love that you've built it on research, and that's what I loved about Tiffanie's work too with EL, tell me how you're at what you're accessing. Are you able to access paywall journals? I know that's always a concern for the field, so it's probably a good question to answer here, and then I have kind of a classroom practical question to ask.
• 11:48
  David Blumenthal: Oh, okay, great, yeah. So the the starting point right now is we're in the proof of concept phase, and we started with what was the easiest way to get a big database together, of good, you know rigorous research that didn't have to require me to read through 400,000 articles?
• 12:05
  David Blumenthal: And so that's where I went to the education resource information center, better known as ERIC, which is a library that has been around since I think, 1965 and was included in the original Elementary and Secondary Education Act. I also speak policy. And so ERIC has been around for a long time and has criteria for what gets included, and we continue to expand that database to include the What Works Clearinghouse and the practice guides that are there that are such great content for helping educators synthesize research and then create practical advice and guidance. We're also incorporating other accessible databases, like there's the Scale Institute at Stanford University, which does a lot of research on AI, there's a lot of questions around AI, and so we want to make sure that we're answering those. Our business model, as we move forward and actually have a functioning business, is that we want to also enter into licensing agreements with the journals to go to you know, AERA, American Education Research Association and all their journals, and then create licensing so that we're getting around those paywalls for our users and then giving them access to the full text of the articles. So for now, we have this proof of concept that exists that's based off of ERIC and What Works Clearinghouse and other vetted resources. And then over time, we want to make sure it's evergreen and always being refreshed with the latest research and evaluation.
• 13:27
  Lisa Dieker: Got it well, and I'll just make a shameless plug too, for something new. I run or I edit the Focus on Exceptional Children. Maggie's on the board.
• 13:33
  Lisa Dieker: It's been in the field since 1969 it is now open source, owned by the University of Kansas, and we now have a consortium that I put in the chat there for you, the openseajournals.com that are all the open source journals in the field of special ed, partnering together to provide resources that have all been vetted. So I do think it's an interesting time to kind of start to harvest to some of those open source pieces. So so I'm going to ask you, kind of move a little bit to this, you know, really practical question. So, as you know, we hear our listeners tell us all the time. You know, I call it tech lash. You know, you've heard a backlash. What was this tech lash like? One more AI tool. So help us understand, yeah, how exactly I'm a teacher that Eddie acts as a personal teaching. What makes it different than if I were just to prompt a generic AI model, like, what will Eddie offer me that that just a general, if I'm a good prompter, would get me so that our teachers will understand?
• 14:30
  David Blumenthal: Yeah, one thing that the the generic chatbots, your ChatGPT, Claude, etc. They're Gemini. They are looking at almost everything that's out there on the internet, and they're not necessarily looking at the full text of what they're linking. So they're finding summaries, abstracts, things of that nature, and then using that to provide responses. What Eddie does is we have downloaded the full text of all the articles, so we have the appendices, which oftentimes. You know, this is super nerdy stuff, but, you know, like the appendices of these research articles often include all of the things that the researchers put together to actually study the intervention or study the program. So that'll be things like rubrics, which are so helpful for classroom teachers, and it'll include more description about what was actually studied, because a lot of the research articles, they don't really talk about what happened. They talk about methodology, which, if you're really into math, and you can read the methodology section, but it's not really practical for educators. So that's where Eddie is a is, is a different kind of system, because we have access to all of that stuff that's so crucial to understand and why it's able to deliver high quality recommend recommendations and create high quality instructional materials based off of what the evidence base has to say. And so, Lisa, you said you had, like, a practical challenge you wanted to.
• 15:58
  Lisa Dieker: Yeah, that was my question. So, you know, I'm struggling with a kid, do I just put it in there? And what would come out different if I put it in my Google Gemini versus Eddie, and it sounds like it would be more evidence based and more specific, maybe. Is that a fair statement?
• 16:14
  David Blumenthal: Absolutely.
• 16:14
  David Blumenthal: Yeah. And so Tiffanie and I were talking about this too, as we were discussing the tool that she built and with EL or EL, and then we did with Eddie, you know, if you went to ChatGPT, or even some of this education-specific chatbots and large language models and AI tools, and you put in a question like, you know, hey, a kid threw a desk at me, what should I do in response to this situation?
• 16:39
  David Blumenthal: Some of the responses that we see when we go and do testing of our competitors and testing of the generic ones, will say, put the kid in a restraint.
• 16:44
  David Blumenthal: Obviously, you don't want to do that. And so when we do the same tests on Eddie, we're finding, again, peer-reviewed journal articles, and they're getting practical recommendations for de-escalation and approaches that are pedagogically sound when it comes to actually instructional decisions that are made. I think that's one of the main distinguishing features that you're going to find, is that you're not going to see hallucinations. You're not going to see disproven practices or things that go against the what the latest research has to say and the latest approaches have to say.
• 17:16
  Lisa Dieker: Perfect. Thank you.
• 17:18
  Maggie Mosher: I have one quick follow up question, and then I'm done with my questions. Lisa and I just got back from Google Cloud Next it was a great conference, but we're seeing a massive shift to Agentic AI. And you had mentioned that yours is built off of agentic AI, and that executes. So when we're talking about agentic AI and having different modules built on top of each other, where is the information we're putting into the chat bot? Where does that go?
• 17:41
  David Blumenthal: Yeah, right now, as we built out this like proof of concept version of Eddie, we wanted to make sure that we were keeping information in a secure way. And we're a small team. It's just me and my co founder, Matt, and so we don't have a large budget, and so we found that the easiest way to deal with that is that Eddie essentially forgets all prior conversations that it had. And so if you engage in a new in a new chat session, it's almost like Eddie completely forgot all your previous chats. That's obviously not a feature that is a compelling feature for the market. And so the new version Eddie Pro, and this will be in the free version as well that we offer. We launched that in June, it'll be able to hold on to those chats, and you can search through the chats and reengage and pick them up. Like, let's say, you know, you had a question about this kid that was, like, really acting out in the class, and you wanted to, you've tried a few things because Eddie gave you some suggestions, and you wanted to follow through on them, and now you're coming back to be like, okay, I tried this, but it worked a little bit, but not that. Well, let's talk about it again, and so that'll be a feature that's built into that.
• 18:42
  David Blumenthal: To get there, we had to do a lot in order to make sure that we're securing the chat history while we are a teacher-facing tool. We don't collect student data in our tool. At the moment, anything that teachers put into a large language model, you're disclosing sensitive information in some depending on what you're saying, and so that's how we're solving it now, is just forgetting the conversation, so we don't have to worry about storing it in the cloud in any way. But we're building in that FERPA compliance and COPA and SOPA and the other acronyms that are out there for data security purposes to ensure that when you are putting in information about a specific student, or yourself as a teacher, that it's stored in a way that is going to be meeting the highest security requirements. And this is also something that you know your your listeners should also pay attention to, is be really careful about what you're disclosing to brand name, large language models, or even the school-specific versions. If you don't know the steps, your.
• 19:47
  Maggie Mosher: No PII, Personal Identifiable Information cannot go in.
• 19:51
  David Blumenthal: Yeah, exactly.
• 19:51
  David Blumenthal: Be very careful about that, because it's a huge risk, and you're ultimately disclosing information to a cloud somewhere. We don't know where that cloud, that cloud server is based, and and so that's a real concern. We're addressing that right now as we build out this new version. For the time being, you can download the chat and store it locally, so whatever protocols you're taking to hold on to information, you could do that and then reupload it and reengage in the conversation.
• 20:19
  David Blumenthal: That's kind of a clunky way of handling that problem, but it makes sure that we're clear on our end that we're not disclosing any PII and then it becomes paramount for the users to be very careful about the steps that they take.
• 20:34
  Maggie Mosher: Well your next version is that mean that you are going to keep it on the local server, or are you going to store it somewhere else in your next version? Have you guys decided?
• 20:43
  David Blumenthal: The next versionm I'll have more clear documentation of what the protocols look like for storing sensitive information. But you know right now, we're planning on making sure that we've got the free version is always going to be available, so you can always ask Eddie questions about the research and get answers and get citations and get on with your day, and then the pro version to create documents and generate instructional materials that will be launching June 23rd is the date that we're targeting for that.
• 21:12
  Lisa Dieker: Great and my my last question, and it's a comment and a question. So we just had a guest here from the Czech Republic on a Fulbright exchange for a week here at KU and her work is Edge of Stories.cz that's Czech Republic, comes in English. And I think it's fascinating that her work has all been getting like 2000 case studies from real teachers and getting AI to give accurate answers that humans approve of. And I just see this world of blending those you know, boots on the ground, case stories and Eddie's research together. So just in the last minute to wrap us up, what would you say you your vision is for the future of teachers having an AI agent involved in their classroom? What would you paint the picture for our listeners?
• 21:58
  David Blumenthal: I think I've been giving this a lot of thought, so I've delved into AI, so I have a lot to say, but I'll be very brief, because we only have about a minute left, in general, you know? So I mentioned Stanford earlier. They put up this assessment, or a synthesis of research on where's the research and what does the evidence based look like for AI in both student facing and teacher facing applications right now, and they looked at everything that had been published that they could find that had been published as of October 2025, and right now we know that teachers are using AI tools to save time, and they're shifting their time to work on other things, hopefully better things that are more things that they can do, that no other system can do. But what we don't know, and which is the which is the frontier here, is what, what kind of improvements happen in the classroom, if anything, for those teachers that are using AI, we don't know if teachers using AI actually improving educational outcomes for their students or for the school overall. So my vision is to be the first company that conducts a rigorous evaluation where we're not just helping deliver evidence, but we're also generating that evidence to help us understand where and how teacher facing applications of artificial intelligence can really push the needle and lead to those better outcomes that our students so desperately deserve.
• 23:24
  Lisa Dieker: Great.
• 23:24
  Maggie Mosher: Thank you, David, so much. That was a perfect answer, and we're grateful that you are here to save our teachers time and money, and we look forward to trying out.
• 23:32
  Maggie Mosher: Eddie.
• 23:33
  David Blumenthal: yeah, thank you so much. Maggie, thank you, Lisa. Eddie AI, E-D-D-I, E-A-I.net there's other Eddie's.
• 23:40
  David Blumenthal: I've got a trademark out there with the trademark office, hopefully we'll be able to fix in the future, but Eddieai.net is the place to go to find the one that we're talking about today.
• 23:49
  Lisa Dieker: Great. Thank you.