Being able to accurately assess what, how, and when students learn is a critical component for thriving educational settings. For elementary and middle school learners and teachers in the United States, the assessment experience typically involves completing multiple-choice questions, true/false statements, and fill-in-the-blank exercises and a lengthy review and grading phase after the fact. In early learning, high stakes observational assessments may result in mis-labeling children’s abilities, having a profound impact on their future learning pathways. These traditional methods do not provide a comprehensive understanding of students' knowledge, skills, and abilities. Such tests don’t take into account the cultural, language, and learning differences among students, contributing to inequitable learning outcomes across racially and economically diverse populations. The testing experience can be intimidating to learners, further complicating the efficacy of traditional classroom assessments. (1)

There is an incredible opportunity to innovate assessment practices in US classrooms to better meet the diverse needs of learners. Technology-enabled tools can assess student learning in more authentic and nuanced ways. Additionally, the use of such tools can save assessment delivery time for educators and present new ways for learners to receive real-time actionable feedback. Artificial Intelligence (AI) is a particularly useful technology for assessment, because of its ability to track massive amounts of data, synthesize and analyze these data, and share insights that both the learner and teacher can engage with. AI-enabled assessments can help educators measure what students do, not just what they say (or write, or think) they can do. They also hold potential to be powerful learning experiences in and of themselves, allowing learners more latitude to explore and engage throughout the assessment. (2)

Even with the opportunity to increase learner engagement and confidence, support continuous feedback, and reduce burdens on educators, the idea of implementing AI-enabled assessments in US educational settings raises many ethical and equity concerns that technology designers, policymakers, and education leaders must consider. The responsible use of AI in assessment must include ways to combat algorithmic bias, ensure privacy and data security, allow for rigorous efficacy research, and ensure equitable access for all learners. Moreover, human oversight and input remains essential, meaning educators and administrators will need support and resources as they implement new assessment practices.

AI-enabled assessment is a very new space and one where new ideas and new innovators are needed. We do not expect fully developed and tested solutions to surface (but welcome those that exist!). This challenge welcomes experimental thinking, and we don’t expect applicants to have everything figured out in order to apply and be selected. Solutions led by individuals and teams who aren’t familiar with product development, for example, but are well-versed in AI capabilities are encouraged to apply.

This Challenge seeks innovative solutions that have strong potential to be pilot tested in US schools serving priority learners in Pre-K - Grade 8. Solutions must benefit all students but should prioritize strategies that support those who face the biggest barriers to opportunity, including Black and Latino learners and all learners experiencing poverty. We’re seeking AI-enabled solutions that improve the quality and equity of assessments by doing one or a combination of the following:

• Analyze complex cognitive domains—such as creativity, collaboration, argumentation, inquiry, design, and self-regulation—rather than just skills-based items   

• Provide continuous feedback that is more personalized to learners and teachers, while highlighting both strengths and areas for growth based on individual learner profiles

• Encourage student engagement and boost their confidence, for example by including playful elements and providing multiple ‘trial and error’ opportunities

Use Cases

To help contextualize this challenge and the kinds of solutions we’re looking for, we’ve developed two specific situations where solutions could be most useful:

• Use Case #1: solution that can be used in an existing assessment structure, but tailor assessment items to learner interests, curricular topics, and learners’ cultural context.

• Use Case #2: solution that scores open-ended responses for key quality features (across multiple areas like subject knowledge, vocabulary, creativity, innovation, etc.). The solution should be able to analyze responses in writing or submitted as transcripts of speech.