After NATS: What 742 Interactions Showed About Testing AI on Science Standards

A UNO session and all-day exhibit at the Nebraska Association of Teachers of Science Fall Conference gave educators four AI systems to question and test, and at least 10 people from different school districts have since asked to learn more about the simulations.

At a touchscreen in a hallway at Omaha’s Henry Doorly Zoo and Aquarium, a conference attendee worked through one of four artificial intelligence (AI) labs while two others stood beside her and watched. That was the idea behind the University of Nebraska at Omaha (UNO) team’s day at the Nebraska Association of Teachers of Science (NATS) Fall Conference on Oct. 9. The team brought AI systems people could question and test, not just hear about.

Three days later, the team’s reach report put numbers to that activity. The four labs logged 742 interactions from an estimated 64 to 76 people on and after Oct. 9. An interaction is one click in the app, and one person can make many, so 742 counts actions, not people.

The people estimate combines head counts at the conference session and exhibit with visitors to the labs’ website after the event. Of those 64 to 76 people, 22 used the labs at home afterward.

What could people actually do?

All four labs ran from one website, so the same tools worked on the exhibit’s touchscreen, on the session’s big screen, and on a phone.

In Ask AI, participants put one question about Nebraska’s science standards to multiple AI assistants and compared the answers.

In Build a simulation, they selected an indicator, one of the specific learning expectations within the state standards, and watched AI turn it into an interactive simulation.

In Find misconceptions, they chose an indicator they teach and used AI to surface misconceptions students may bring to it.

In Interview a student, participants played the student while an AI interviewer asked questions about the learning outcomes.

Taken together, the four labs approached a lesson from four directions: what AI says about a standard, how an idea can be represented in an interactive simulation, where students may encounter misconceptions, and what it feels like to be questioned about the material.

The material behind the labs was built around Nebraska science standards. Each of the 326 indicators in the state’s K-12 science standards had an associated misconception quiz and student interview pack. The simulation library included 58 simulations for grades 3 through 12, including 39 set in Nebraska contexts.

From the session to the exhibit

Mahadevan Subramaniam, Ph.D., director of UNO’s Artificial Intelligence Center for Collaborative Outreach, Research, and Education (AI-CCORE), and Parvathi Chundi, Ph.D., professor of computer science, presented “Demystifying AI” with AI-CCORE engineer Vishva Prasanth Srinivasan.

The 50-minute session was designed as a hands-on exploration of large language models, guardrails, and classroom applications. The team ran the labs live on the room’s big screen, and attendees could follow along on their phones. About 12 people attended the session.

The session and exhibit were designed to work together. The labs’ website also connected attendees with the exhibit run through the Nebraska AI Literacy and Workforce Infrastructure Initiative (NA’WI), which AI-CCORE leads at UNO with Nebraska EPSCoR.

The exhibit gave people additional time to try the labs at their own pace and talk with team members. An estimated 30 to 42 people engaged with the three exhibit stations.

Before the event, the team also tested how the systems behaved.

Across 96 answers from seven AI models working without Nebraska’s standards document, none named the correct Nebraska standard code. In 10 attempts to get the AI interviewer to give away answers, it redirected all 10. A second AI model also checked all 652 quizzes and interview packs.

That first result reflects one of the questions behind the Ask AI lab: does an AI assistant actually know the answer, or does it simply sound sure?

After the exhibit closed

The labs did not disappear when the event ended. They remained available online for people to use on their own devices.

Twenty-two people used the labs at home after the event. They are part of the estimated 64 to 76 total, not additional to it, and represent roughly a third of everyone counted.

Because this group was measured through visits to the website, the report cannot establish whether those 22 people had first used the labs at the session or exhibit. What the data does show is that the labs were used away from the conference setting, on people’s own devices.

A second kind of follow-up came afterward. Srinivasan said at least 10 people from different school districts have reached out since the event to learn more about the simulations.

An inquiry is not adoption, but it does show that questions about using the simulations continued after the event.

What the numbers show, and what they do not

People spent time testing AI systems against Nebraska science standards at the conference and away from it. Some used the labs at home afterward. People from different school districts also contacted the team to learn more about the simulations.

That behavior matters when thinking about AI literacy. One practical way to evaluate an AI tool is to test it using material you already understand, and the labs gave participants an opportunity to do that with the standards Nebraska science teachers already use.

What it took, and what comes next

Getting the labs in front of people required physical work as well as software. The team transported a large touchscreen kiosk to Omaha’s Henry Doorly Zoo and Aquarium, assembled the exhibit, and brought the digital experience into the conference setting.

One practical lesson also came from the traffic data. Of 90 page views recorded in the report, 70 came from phones. That means the experience on a phone matters alongside what participants see on a large exhibit screen.

The four labs remain available at:

nats.aiccore-uno.ai

Educators and school districts interested in learning more can also message AI-CCORE on LinkedIn.