Aula de cursos

Goal: use dialectical thinking for decisions with multiple valid but conflicting priorities, without handing the decision to a chatbot.

1. A new phone you might not need

Your battery dies before the end of the day. An AI instantly recommends ten new smartphones. But the actual decision includes battery repair, price, longevity, warranty, accessibility, essential features and what spending this money would displace. “Buy or put up with it” may be a false dilemma. Compare new purchase, repair, used devices and postponement under the same requirements; verify local specifications and costs.

2. A higher-paying job that costs more of your life

Another position offers a larger salary but a significantly longer commute. Ask the model to compare income after relevant costs, commuting time, professional learning, uncertainty and family responsibilities. Your preference for time with loved ones is a legitimate value; it does not need statistical proof. The assistant should not supply values you never expressed.

3. When every source is genuine but the recommendation is still wrong

Imagine asking: “I am exhausted by my job. Should I quit and start a business?” The AI cites authentic studies on work satisfaction, trustworthy statistics on entrepreneurship and success stories. It ends with a stirring invitation to live boldly. Every reference might be real, yet none establishes that this particular person can pay rent in three months. The model may not know existing debts, dependents, health coverage, cash reserves, the local job market, or whether the business has any paying customers.

That is an important distinction: a factual hallucination invents or misstates a fact; an unsupported inference reaches farther than its evidence allows; a selection bias highlights successes while omitting failures; an imprudent recommendation ignores foreseeable consequences. Checking that citations exist only catches part of the problem.

4. The dialectical issue: freedom that could undermine freedom

Leaving a job may free up time and create room for independent work. But losing dependable income could reduce the material autonomy necessary to use that freedom. Remaining without change might also be harmful. The lesson is not to label resignation heroic or cowardly; it is to inspect which conditions make the choice sustainable and what information remains absent.

5. Laboratory: audit a life-changing recommendation

1. Ask an AI for advice on a fictional resignation and entrepreneurship scenario, requesting traceable studies if its tools permit.
2. Separate population-level findings from claims about the individual. Check survivorship bias, omitted negative outcomes and assumptions about finances.
3. Add explicitly fictional obligations, savings, living costs and current business revenue.
4. Compare reversible options: testing demand while employed, seeking a new job, negotiating workload, or creating a conditional transition.
5. Request a conservative cash-flow projection with downside cases. Verify its arithmetic and clearly mark fabricated inputs.

Do not persuade me to quit or stay. Audit your prior advice. Separate verified facts, population-level findings, individual inferences and value judgments. Identify missing financial conditions, omitted downside outcomes and what would reverse your recommendation. Include reversible alternatives and evidence to obtain before deciding. If a responsible recommendation is not possible, say so.

Submission: a four-column table: authentic source; what it actually supports; what it does not support; individual evidence still required. The responsible answer may be to withhold a recommendation. Do not provide confidential personal data to an external service.