I remember the first tech tool I used in my classroom as a high school math teacher that I did not use as a math student: the Desmos graphing calculator. Its utility and the implications were instantly clear. Graphs were more colorful and precise. There was no need for small warehouses worth of AAA batteries. Students could use Desmos on a laptop, desktop, or even on their phone, and the platform looked largely the same as anything I projected for my whole class to see. I could even share pre-made graphs with my classes with a hyperlink, instead of going one-by-one to each handheld calculator with a connecting cord. All these features were particularly helpful for me as a traveling teacher, without my own room to store anything in, in a school where students also had no lockers.
The Desmos graphing calculator is still alive and well. Now it is embedded or allowed on several nationally administered exams and assessments, and still freely available to anyone like it was when I first used it with my students. However, edtech stories like this are more the exception than the rule. Many ideas and products that promised evolution, or even revolution, have come and gone with little impact.
Artificial Intelligence (AI) erupted in the last couple years and seems certain to influence education. Predictably, many edtech products featuring AI have popped up, and in all likelihood most of them will not impact education. Colleague AI is looking to be the exception through a teacher-centered design process. If you are someone like me, who tends to be skeptical of edtech products, then please allow me to share more about Colleague AI and why it is more likely to be the kind of impactful edtech product often promised but rarely delivered.
Colleague AI’s Development Process
Colleague’s design combines the methodologies of Human-Centered Design Thinking with IBM’s Enterprise Design Thinking to empathize with users, define design goals, ideate solutions, prototype, and test solutions in a fluid approach. The result is an AI product focused on augmenting teacher expertise with researched best practices.
Colleague’s design process started in 2021 with interviews and surveys of over 500 educators nationwide to understand lesson planning experiences, types of instructional materials used daily, experiences with edtech, and identify common dilemmas. Colleague analyzed interview responses and then conducted competitor analysis to understand existing tools’ strengths. Lesson planning and lesson adjusting emerged as significant areas ripe for improved tools with the potential to meaningfully benefit teachers and classrooms.
Next, Colleague conducted co-design sessions with 10 teachers across multiple states, grades, and subjects to ensure the broader utility of the platform. This involved interactive brainstorming sessions and led to the creation of the first low-fidelity prototype. Teachers were re-invited to evaluate the prototype and revisions led to the mid-fidelity prototype. From there, Colleague’s development focused on math by working with 15 math educators to evaluate the usability and feasibility of the mid-fidelity prototype. Feedback was used to develop the high-fidelity prototype, which further evolved after working with the software engineering team to ensure technological feasibility and scalability.
Nearly all teachers who experienced the prototyping phases praised Colleague’s intuitive design, and Colleague continues to seek out feedback from active teachers. They recruited an educator panel that represents of the teacher workforce in terms of demographic diversity, years of experience, grade level, and types of students served. Teachers are currently annotating more data, identifying bugs, offering feedback on useability, and proposing desired features. Teacher perspectives and expertise will continue to be the backbone of Colleague AI’s model, and the engine that propels future developments and evolution of the product.
The Human-Centered Difference
Colleague’s design and development, employing a human-centered AI partnership model, gives it a distinct edge over its competitors in many ways. For starters, it became obvious early on in interviewing teachers that pedagogical approaches are distinct for different grade levels and different academic subjects. So, in response, Colleague continues to hone specialized AI algorithms and models for a given subject of instruction (current subjects supported are math, science, and ELA.) Each algorithm is developed by integrating advanced AI with research knowledge and human expertise in each subject. This allows Colleague AI to more seamlessly integrate various relevant inputs, such as the features of lesson materials, teaching styles, and classroom considerations (like multi-language learners, special education accommodations, and neighborhood contexts), and generate personalized instructional materials.
Another trend noticed in teacher interviews is that teachers often use multiple tools and sources when lesson planning. So, Colleague has invested substantial time in building a platform with features that eliminate the need to jump around to create a lesson. These features include standards-aligned lesson generation, vetted open education resources, student formative assessment generation, rubrics-based grading, file editing, and integration with several major LMSs via web browser extension. Colleague harnesses the power of AI to offer dynamic material generation and recommendation, significantly diminishing teachers’ time invested in material vetting, selection, and creation. This saves teachers the time it takes to determine the quality of lesson from popular online sources, such TeachersPayTeachers and social media platforms, where materials have no quality vetting. Use of Colleague AI also outperforms traditional open education resources (OER) platforms (e.g., Illustrative Math, BetterLesson, Achieve the Core) in efficiency thanks to embedded features in the platform, in particular the ability to edit existing OER lessons.
Teachers as Change Agents
The future remains unwritten, but Colleague’s belief is that AI will unleash the potential for teachers to transform learning in their classrooms. As a result, Colleague is designed with teachers at the center and sees growth of the product inherently tied to teacher use. Teachers already look to constantly improve the curricula they use and their instructional methods, and Colleague AI looks to provide teachers new ways to do this.
Colleague’s lesson materials will rapidly increase in quantity and reflect the most updated understanding of effective instruction in the AI era, disrupting traditional textbooks. The rapid development of multi-modal generation models will unleash teachers’ creativity in generating images, texts, and videos to engage their students in learning complex mathematical concepts and skills. Teachers’ use of Colleague will make the algorithms smarter, capable of generating better and unlimited variation of lesson materials, personalized to individual teachers’ and students’ needs. More importantly, Colleague’s ongoing human-centered approach, with teachers and experts at the heart of development, enables Colleague to continually embed the latest research into its algorithms and design principles.
Colleague also offers several types of feedback to teachers, including ratings of many dimensions of lesson materials, prompts for refinement generated by nudging algorithms, student diagnostic reports, and recommendations of materials and instructional strategies. The “lesson simulation” feature can help teachers anticipate how their lessons may unfold receive AI reviewer’s feedback before they teach it. These features empower teachers to work on their own practice in ways that are typically supported through activities like learning walks and instructional coaching cycles. While these kinds of professional learning opportunities are often highly impactful, but not always accessible to teachers due to a variety of limitations, such as a lack of substitute teacher availability or a lack of an instructional coach in a building. Colleague’s basic functions are offered for free to teachers to democratize the access to quality instruction and the potential benefits in classrooms. Wider access also helps mitigate algorithm bias due to limited training data and user bases. This is yet another example of how Colleague’s human-centered approach, focused on improving human interactions in the classroom, results in win-win situations where benefitting teachers simultaneously strengthens the product.
Final Thoughts
Colleague started with teachers in mind and continues to put the demands of the teaching profession at the center of Colleague AI’s development. To go back to the Desmos example I started with, I already had experience working with a graphing calculator. Learning about Desmos was intuitive, even when accessing a web-based product instead of grabbing a device from my desk drawer felt unnatural.
Implementing AI into my own practice feels like a big leap, but like nearly any teacher, I constantly think about lessons and instructional moves to try to meet the needs of every student. I do my best to anticipate student responses as part of lesson planning. These are not new activities, but Colleague AI provides new ways to do these more efficiently, comprehensively, and deeply. It is not the only AI tool that provides this opportunity, but their exceptional design process leadd to a product uniquely qualified for the task.
