# How to design academic activities that incorporate artificial intelligence

> Learn how to design academic activities with artificial intelligence that preserve evidence of learning, foster critical thinking and set clear rules.

- Site: Genialoh (https://genialoh.org)
- Language: en
- Category: Teaching
- Reading time: 14 min
- HTML version: https://genialoh.org/#/en/blog/designing-academic-activities-with-artificial-intelligence

---
Artificial intelligence can be incorporated into an academic activity in many ways: to explore ideas, compare answers, explain concepts, analyze information, practice procedures or review work.

However, adding an AI tool to an existing activity does not guarantee it will improve learning.

A well-designed activity must preserve its academic purpose. It must also define what the tool can do, what decisions belong to the student, how information will be verified, and what evidence will allow learning to be assessed.

> **Change the question**
> 
> The main question should not be: how can I use AI in this activity? The more useful question is: what does the student need to learn and what role can AI play without replacing that learning?

This guide presents a practical process to design academic activities that incorporate AI intentionally, transparently and responsibly.

## What is an academic activity with artificial intelligence?

It is an activity in which AI has an explicit role within the learning process.

- Generate a first proposal or create examples.
- Explain a concept or formulate questions.
- Compare alternatives or analyze information.
- Identify errors or simulate a situation.
- Propose feedback or review clarity.
- Support practice or explore possible solutions.

The activity should not be limited to asking the student to obtain an answer from the tool. It must require them to interpret, verify, apply, modify, question or defend the result.

## The risk of adding AI without redesigning the activity

An activity created before AI became widespread can easily be delegated. For example: "Write a 1,500-word essay on leadership." If the instructor only receives the final product, it will be hard to know what the student understood, what decisions they made, what sources they consulted, what parts were generated, what they verified, what errors they identified and what they contributed personally.

The answer does not have to be automatically banning AI. The activity can be redesigned so the tool participates without hiding evidence of learning.

| Original activity | Redesigned activity |
| --- | --- |
| Write an essay on the effects of AI on employment. | Select an economic sector and formulate a specific question. |
|  | Ask AI for two different explanations and identify claims that require evidence. |
|  | Verify those claims with academic or institutional sources. |
|  | Detect errors, omissions or generalizations. |
|  | Write your own argument and explain what you accepted or rejected. |
|  | Submit a use disclosure and defend your conclusions orally. |

The tool participates, but the student remains responsible for researching, comparing, verifying and arguing.

## Step 1. Define the learning objective

Every activity should start from the result you expect to observe. The instructor can ask what the student must understand, what skill they must demonstrate, what decision they must be able to make, what procedure they must perform, what product they must create, what professional criterion they must apply, and what evidence will show they learned.

| Vague objective | Observable objective |
| --- | --- |
| That the student knows about AI. | That the student compares two AI-generated answers, identifies claims that need verification and builds a conclusion supported by sources. |

### Useful verbs for AI activities

Use actions that require intellectual participation: analyze, compare, classify, verify, justify, critique, correct, apply, design, evaluate, defend, interpret, improve, document and reflect. Verbs like "generate" or "obtain" can be part of the activity, but they should rarely represent all expected learning.

## Step 2. Decide what role AI will play

The tool must have a bounded role.

| Role | What AI does | What the student does |
| --- | --- | --- |
| Alternative generator | Proposes hypotheses, examples, structures, questions or scenarios. | Selects and justifies the most suitable alternatives. |
| Object of analysis | Produces an answer to examine. | Identifies errors, biases, verifies claims and compares perspectives. |
| Practice tutor | Explains concepts, formulates questions, proposes exercises and gives hints. | Practices, solves and knows when to seek a human. |
| Simulator | Represents a client, fictional patient, interviewer or committee. | Interacts with the simulation, aware it is not a real person. |
| Review assistant | Reviews clarity, organization, grammar and omissions. | Decides which changes to accept and retains responsibility for the final version. |

## Step 3. Set the allowed use level

Every activity must clearly communicate what AI use is authorized.

| Level | When it applies | Sample instruction |
| --- | --- | --- |
| Not allowed | Evaluate individual knowledge, spontaneous writing or performance under controlled conditions. | Clarify whether spell checkers or calculators are allowed. |
| Limited | Use only in certain phases. | You may use AI to generate study questions; not to solve the case or write the conclusion. |
| Allowed with disclosure | The student may use it, explaining how. | Document tool, purpose, stage, prompts, results incorporated and verification. |
| Required | The tool is part of the learning. | Generate two answers with different prompts, compare them and explain how the context changed the result. |

There should be no ambiguity. The rule must appear inside the activity's instructions.

## Step 4. Design a task that does not end with the generated answer

The AI result must be a starting point or a work element. The activity can ask the student to identify errors, verify data, compare with sources, rewrite the result, apply the proposal to a real case, defend a decision, explain limitations, document iterations, classify suggestions, reject inappropriate answers or generate their own alternative.

> **Useful pedagogical formula**
> 
> Generate + analyze + verify + produce + reflect. The activity does not depend on the AI producing a perfect answer: even a problematic answer can become learning material.

## Step 5. Preserve evidence of the process

When only the final product is evaluated, it is hard to observe how the student participated. The activity can request drafts, version history, decision journal, sources consulted, prompts used, relevant generated responses, change table, final reflection, oral explanation, presentation, follow-up questions or practical demonstration.

It is not necessary to ask for all of these in every activity. Choose the evidence that matches the objective and risk level.

### Sample short journal

| Stage | AI use | Student decision |
| --- | --- | --- |
| Planning | Generated three possible structures. | Chose the second and changed the order. |
| Research | Suggested sources and concepts. | Discarded references I could not verify. |
| Drafting | Reviewed the clarity of a paragraph. | Accepted two changes and rejected one. |
| Closing | Proposed a conclusion. | Wrote a different conclusion. |

## Step 6. Require verification

AI can produce incorrect claims, nonexistent references, wrong calculations, faulty code or incomplete interpretations. The activity must specify how the result will be verified.

- Consult academic sources or official documents.
- Check calculations or run the code.
- Contrast with a standard or with real data.
- Review the original source and validate with experimental evidence.
- Explain a procedure and identify the degree of uncertainty.

> **Recommended instruction**
> 
> Do not present as true any AI-generated claim without verifying it through the sources or procedures defined in this activity.

## Step 7. Incorporate academic context

Answers improve when the activity provides clear context: student level, objective, program, methodology, case, criteria, allowed sources, restrictions, audience and format.

An institutional platform can keep programs, methodology, values and guidelines provided by the university present. This allows activities to be designed around tutors configured by degree or program, rather than relying solely on a general tool without institutional context. Context does not eliminate errors: verification and human oversight remain necessary.

## Step 8. Protect information

> **Do not enter into tools**
> 
> - Names, IDs and student records.
> - Individual grades.
> - Medical, financial or identifiable personal information.
> - Passwords and confidential assessments.
> - Unpublished question banks and reserved research.
> - Confidential institutional documents or protected materials.

When real cases, works or examples are used, they must be anonymized. The university should communicate which platforms are authorized and what conditions apply to institutional information.

## Step 9. Define how use will be disclosed

> **Short disclosure**
> 
> I used [tool] for [purpose]. I reviewed and modified the result as follows: [explanation]. I verified the information through [sources]. I take responsibility for the final content.

The length should be proportional. A brief activity may need a three-line disclosure. A final project may justify a more complete journal with tool, purpose, main prompts, incorporated result, modifications, verification and limitations.

## Step 10. Create appropriate assessment criteria

An activity with AI should not be assessed solely on the visual or verbal quality of the final product. The rubric can include understanding of the problem, analysis, verification, use of sources, justification, quality of the student's own product, transparency and ability to explain.

| Criterion | Excellent | Satisfactory | Developing | Insufficient |
| --- | --- | --- | --- | --- |
| Analysis | Deeply evaluates and detects relevant issues. | Analyzes main aspects. | Partial analysis. | Accepts result without analysis. |
| Verification | Checks with appropriate sources. | Verifies most. | Limited verification. | Does not verify. |
| Decisions | Justifies what was accepted or rejected. | Justifies main decisions. | Little explanation. | Does not explain decisions. |
| Own product | Integrates evidence and a solid proposal. | Adequate answer. | Too dependent on AI. | Presents generated result as own. |
| Disclosure | Precisely documents use. | Includes essential information. | Incomplete disclosure. | Does not disclose use. |

The rubric should be shared before starting the activity.

## 15 types of academic activities with artificial intelligence

| # | Type | What the student does |
| --- | --- | --- |
| 1 | Comparison of explanations | Generates two explanations for different audiences, verifies concepts and creates their own version. |
| 2 | Error detection | Identifies, classifies and corrects errors in a generated response; proposes a better one. |
| 3 | Reference verification | Checks which references exist, which are relevant and which data are correct. |
| 4 | Debate with a generated response | Builds counterarguments and defends their position orally. |
| 5 | Professional simulation | Asks questions to a fictional client, makes decisions and documents limitations. |
| 6 | Iterative product improvement | Creates an initial version without AI and decides which feedback to accept. |
| 7 | Generic vs. institutional tutor | Asks the same question in two contexts and analyzes differences. |
| 8 | Case design | Adapts an AI-proposed case to a local context, industry or regulation. |
| 9 | Bias analysis | Modifies variables and examines differences, assumptions and mitigation. |
| 10 | Socratic tutoring | Logs questions, hints received and changes in reasoning. |
| 11 | Proposal evaluation | Evaluates a generated strategy with defined criteria and produces a corrected version. |
| 12 | Explanations for different audiences | Generates versions for different profiles and identifies what changes. |
| 13 | Code analysis | Runs it, detects errors, tests edge cases and documents risks. |
| 14 | Procedure review | Analyzes assumptions, order, missing steps and alternatives. |
| 15 | Responsible-use portfolio | Documents cases, prompts, results, errors, verifications and decisions. |

## Complete activity example

> **Critical evaluation of an AI-generated recommendation**
> 
> - Objective: evaluate a professional recommendation with disciplinary criteria, evidence and risk analysis.
> - AI use: required.
> - Steps: read the case, formulate a prompt, save the result, identify five key claims, verify them, evaluate with the rubric, identify risks, write your recommendation and submit a disclosure.
> - Evidence: prompt, response, verification table, evaluation, final recommendation, disclosure and a five-minute oral defense.
> - Criteria: understanding, quality of evidence, verification, disciplinary application, justification, reflection and transparency.

## Template to design an activity

| Element | Content |
| --- | --- |
| 1. Name | Clear title related to the task. |
| 2. Learning outcome | What must the student demonstrate? |
| 3. AI role | Generation, explanation, comparison, simulation, tutoring or review. |
| 4. Use level | Not allowed, limited, allowed with disclosure or required. |
| 5. Authorized tool | Institutional platform or approved solution. |
| 6. Instructions | Concrete, sequential steps. |
| 7. Process evidence | What the student must keep and submit. |
| 8. Verification | What sources, tests or procedures to use. |
| 9. Information protection | What data cannot be shared. |
| 10. Disclosure | What information to provide about AI use. |
| 11. Assessment criteria | How process, decisions and product will be evaluated. |
| 12. Contingency | What to do if the tool fails or is unavailable. |

## How to implement these activities in an institution

A university should not rely solely on the individual initiative of each instructor.

- Define common principles and set use levels.
- Approve tools and create templates.
- Train faculty and design pilot activities.
- Review by discipline and apply in two programs.
- Collect feedback and document examples.
- Adjust guidelines and expand gradually.

Academic coordinators can create a repository of reviewed activities for other instructors to adapt.

## How to measure whether an activity worked

- Understanding of instructions and quality of disclosures.
- Verification capacity and process evidence.
- Quality of products and participation.
- Recurring questions and tool errors.
- Privacy incidents and faculty preparation time.
- Perceived usefulness and training needs.
- Platform changes required.
- Student's ability to explain their work.

The number of interactions does not by itself demonstrate learning. Results must be analyzed alongside academic evidence and faculty observations.

## Common mistakes when designing AI activities

| Mistake | Consequence |
| --- | --- |
| Adding AI without changing the task. | The tool does what the student was supposed to demonstrate. |
| Assessing only the final product. | Process, verification and decisions are not observed. |
| Not communicating rules. | The student does not know what uses are allowed. |
| Asking only for a generated answer. | No analysis or original production is required. |
| Trusting generated references. | They may be incorrect or nonexistent. |
| Not protecting data. | Personal or confidential information is shared. |
| Requiring disproportionate records. | Unnecessary burden. |
| Same activity across disciplines. | Each program needs different criteria and cases. |
| Fully automating assessment. | Human oversight and responsibility are lost. |
| No fallback plan. | No alternative when the tool fails. |

## Frequently asked questions

### Should every activity incorporate artificial intelligence?

No. AI should be used when it adds value to the learning objective. Some skills need practice and evaluation without assistance.

### Must the student submit all their conversations?

Not necessarily. Evidence should be proportional. You can request only the main prompt, the relevant responses and a reflection.

### How do I prevent AI from doing all the work?

Design tasks that require verification, application, decisions, original production, oral defense and process evidence.

### Can AI be used in assessments?

Yes, when the objective and rules justify it. There can also be assessments where it is prohibited.

### What if the tool generates incorrect information?

The student must identify and correct the errors. Verification capacity can be part of the evaluation.

### Can student work be used as examples?

It must be anonymized and used only with the appropriate institutional authorizations and conditions.

### Is it necessary to cite the tool?

The institution should define a common form of disclosure. Academic claims must be supported by verifiable sources.

### How is an AI activity assessed?

Assess understanding, process, verification, decisions, original production and transparency, not just the final presentation.

### Can Genialoh support instructors?

Genialoh includes AI support for instructors, tutors by degree or program, contextualization with syllabi, faculty training, implementation, support and adoption tracking.

### How can a university preview these activities before implementing?

It can select two degrees or programs and prepare a functional demo with its identity, academic context and representative use cases.

## From a generic tool to a distinctive academic experience

The best AI activities are not the ones that produce more content. They are the ones that help the student analyze, verify, apply, decide and reflect. To achieve this, instructors need clear objectives, visible rules, appropriate tools and an experience aligned with the academic program.

> **Genialoh for your institution**
> 
> - Institutional name, logo and colors.
> - Context based on syllabi, programs, values and methodology.
> - Tutors configured by degree or program.
> - AI support for instructors and personalized learning.
> - Institutional reports and analytics.
> - Training, implementation, support and adoption tracking.

Book a first 30-minute meeting. If your institution decides to move forward, it will select two degrees or programs and provide the required authorized academic documentation. Genialoh will prepare a functional demo with the institution's identity and context in less than 14 days from receiving the complete documentation. After the presentation, up to five authorized people can explore it for 30 calendar days.