AI Systems and Tradeoffs

Overview

Students begin a foundations of artificial intelligence course by investigating how an AI model uses data to make predictions and how design choices affect people. Working in pairs, they use a simple computer simulation to test an AI-assisted disaster-response system, then evaluate its accuracy, fairness, cost, safety, and social impact.

Learning intentions

Students will be able to:

• Describe how data, models, predictions, and criteria connect in an AI system.
• Identify criteria and constraints for an AI solution to a real-world problem.
• Use a computer simulation to test how changing model settings affects outcomes.
• Explain tradeoffs and recommend a responsible design choice.

Success criteria

• I can explain the basic pathway from data to model to prediction.
• I can identify at least three criteria and two constraints for an AI disaster-response system.
• I can use simulation evidence to compare two proposed settings.
• I can justify a recommendation using evidence and acknowledge at least one tradeoff.

Curriculum links

• Analyze a major global challenge to specify qualitative and quantitative criteria and constraints that account for societal needs and wants.
• Design a solution to a complex real-world problem by breaking it into manageable engineering problems.
• Use a computer simulation to model how proposed solutions affect interactions within and between systems.
• Evaluate solutions using prioritized criteria and tradeoffs, including cost, safety, reliability, and social and environmental impacts.

Lesson structure (50 minutes)

1. 0–5 min · Hook and course launch. Teacher displays the opening scenario in the AI emergency-response introduction deck: “A city has one hour to send aid after a major storm. Should an AI system decide who receives help first?” Students make a quick individual choice and share one reason with a partner. Teacher explains that this is the first lesson in a foundations of AI course and that no programming experience is expected.

2. 5–12 min · Build the foundation. Teacher uses the AI emergency-response introduction deck to introduce four terms: data, model, prediction, and decision. Teacher explains that an AI system does not “understand” a situation like a person; it identifies patterns in available data and produces an output based on its design. Students annotate the first section of the AI foundations and simulation worksheet by connecting each term to the emergency-response example.

3. 12–18 min · Define the engineering problem. Teacher presents the challenge: design settings for an AI system that prioritizes neighborhoods for emergency supplies after a storm. The system must consider need, travel time, population, road access, and reported damage. In pairs, students use the worksheet to identify criteria, such as lives protected and delivery time, and constraints, such as limited vehicles, fuel, incomplete data, safety, and fairness. Teacher checks that students distinguish a criterion from a constraint.

4. 18–33 min · Run the simulation. Teacher demonstrates the controls in the AI emergency-response introduction deck and assigns pairs one shared computer. Students open the prepared disaster-response simulation and test at least two settings, changing the priority given to need, travel time, and reported damage. They record predicted delivery outcomes, unmet needs, resource use, and any unexpected effects on the AI foundations and simulation worksheet. Teacher circulates, asking: “What did the model prioritize?” and “Who benefited or lost when you changed that setting?”

5. 33–42 min · Evaluate and improve. Teacher pauses the class for a brief comparison using the AI emergency-response introduction deck. Pairs compare results and choose their preferred design. They score it against the class criteria, then identify one tradeoff—for example, faster delivery may reduce fairness, or using more data may increase cost and privacy concerns. Students propose one improvement, such as collecting better data or adding a human review step.

6. 42–50 min · Debrief and missed-class pathway. Teacher leads a discussion: “Can an AI system be neutral if its data or criteria are incomplete?” Students complete the worksheet reflection and exit response: “Our recommended setting is ___ because the simulation showed ___. One risk or limitation is ___.” Teacher directs absent students to use the slide deck, simulation directions, and worksheet independently or with a partner during the next available work period; the worksheet includes a shortened no-device alternative using the sample results table.

Resources

• the AI emergency-response introduction deck
• the AI foundations and simulation worksheet
• Teacher computer and projector
• Ten student computers or tablets, one per pair
• Prepared disaster-response simulation or spreadsheet model
• Internet access if required by the simulation
• Whiteboard and markers
• Pair role cards or a visible assignment of Driver and Analyst
• Printed absent-student directions included with the worksheet

Assessment

• Formatively assess pair definitions of data, model, prediction, criterion, and constraint during the first three activities.
• Check simulation records for evidence that students changed settings, compared outcomes, and identified interactions or tradeoffs.
• Use the exit response to assess whether each student can support a recommendation with simulation evidence and recognize a limitation.

Differentiation

• Provide a word bank, labeled model diagram, sentence starters, and a completed example row for students who need language or organizational support.
• Pair students strategically and assign roles; allow students to use the worksheet’s sample-results alternative if device access, reading, or processing needs make the simulation difficult.
• For multilingual learners, preview terms with visuals and permit oral rehearsal before writing; avoid assuming that unfamiliarity with AI reflects lack of reasoning ability.
• Challenge ready students to propose an additional criterion, explain how it could change the model, and predict a possible unintended consequence before testing it.