A self-paced critical thinking course for any professional who has to judge a supplier’s claim, a dashboard, a forwarded article or an AI summary, and then pass it on to someone who will act. You follow one analyst, Mika, through eighteen weeks and one buying decision, from forwarding a slogan to writing a case that holds. Imagine being the person in the meeting who can say what a number means and how you know. About 11 hours over 56 days.
Bodhih Critical Thinking in the AI Era is a self-paced online course for any professional who must judge claims, numbers and AI answers before passing them on. Eight 75-minute modules follow one analyst through a fictional buying decision and teach how to pin down a claim, weigh sources, catch thinking traps, read misleading numbers, check AI output at the source, test causes, decide under uncertainty and write a reasoned recommendation. Each module has an animated video and a browser simulation. Learners earn a Bodhih credential with an Open Badge.
Facts on this page last checked by Siddharthan VGJ, Director, Bodhih Training.
A slogan is not a claim. The animated concept video from Module 1 (What Is Actually Being Claimed), narrated by a synthetic voice, with captions.
An AI answer, a supplier’s deck or a forwarded article reads well and arrives finished. You send it on. Then someone asks how you know, and you cannot say. Nobody lied to you. The answer was fluent, you were busy, and your name is now on it.
A figure such as “40% better” or “90% accurate” lands in a meeting and nobody asks what it was measured from, or against what. The room agrees fast, the most senior voice carries it, and the doubts come out in the corridor afterwards.
You have been caught, or fear being caught, forwarding a reference, a quote or a figure that an assistant made up. So some days you check nothing and some days you check every line and never finish. What is missing is a rule for what to check first and when to stop.
For people who write the briefings and numbers that others act on. You learn to trace every claim to where it came from, and to state plainly what you did not check.
For people who decide on what a team brings them. You learn how far to trust a fluent page, how to slow a room that agrees too fast, and how to name a checker with a task that could fail.
For people who choose tools and suppliers under time pressure, or who make claims to customers. You learn to turn a headline into a claim, and to redraw a number so that it shows its size.
For teams that want one shared way to check a claim. Picture a team where anyone can ask “compared with what?” and nobody hears it as an attack. Several of the at-work tasks suit a group.
This course is not for you if you want to learn prompting or a particular AI product, formal logic, statistics as a subject, or research methods for scientists: it teaches none of those, and it has no live classes or human tutor.
Five dated statements, each linked to its source and given with its limits. Every source was opened at the original on 14 November 2026. None of them shows that using AI makes anyone think less well, and this course does not claim that.
The US standards body NIST lists “confabulation” among the risks of generative AI: systems that “generate and confidently present erroneous or false content”. It is a voluntary framework for organisations, not a rule for individuals.
National Institute of Standards and Technology, “Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile”, NIST AI 600-1 (read 14 November 2026), July 2024In a Microsoft Research survey of 319 knowledge workers, people who had more confidence in the AI reported doing less critical thinking, and people with more confidence in their own skill reported doing more. The survey is self-reported and shows a link, not a cause.
Lee and colleagues, “The Impact of Generative AI on Critical Thinking”, Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, Microsoft Research (read 14 November 2026), April 2025One major AI builder’s own research explains that models are trained and scored in ways that reward a confident guess over “I do not know”. It is a builder explaining its own technology, so it is an informed account and not a neutral one.
OpenAI, “Why language models hallucinate” (read 14 November 2026), 5 September 2025Another builder’s own documentation recommends checking habits such as permitting “I do not know” and asking for exact quotes, and says that these steps reduce made-up content and do not eliminate it.
Anthropic, “Reduce hallucinations”, Claude Platform documentation (read 14 November 2026), Undated page, as read on 14 November 2026In an experiment, very experienced radiologists rated about 82% of scans correctly when the AI’s suggestion was right, and about 46% when it was wrong. It was a small experiment in one medical task, and the figures do not transfer to other work.
Dratsch and colleagues, “Automation bias in mammography”, Radiology 307(4) (read 14 November 2026), 2023Each module leaves you with one check you can use in a meeting or before you press send, and a one-page job aid for it. Nobody can promise that your decisions will be better, and this course does not. It gives you honest evidence, practice on a realistic case, and tasks on your own work.
Ask five questions (what exactly, compared with what, measured how, over what period, says who) and say what is still unknown.
Leave the page to see who is behind it and who pays, and tell four sources from one source repeated four times.
Notice anchoring, fluency, confirmation and automation bias, and use a specific counter-move instead of “being careful”.
Restate any percentage, risk or accuracy figure as plain counts with its starting point, and spot a chart that exaggerates.
Decide which parts must be checked, check them at the original source, and say plainly what was not checked.
List rival explanations, run a premortem, set a decision rule in advance, and write a one-page recommendation that states the other side fairly.
The pathway runs in order: a baseline assessment, then eight modules (what is actually being claimed; weighing evidence and sources; thinking traps and the fluent answer; numbers that mislead; checking AI output; causes, correlation and rival explanations; deciding under uncertainty; making and testing a case with others), and a final assessment.
Measured before you start and again on the same instrument at the end, so the change is recorded rather than claimed.
The baseline assessment, Form A: 20 scenario questions on all eight competencies, about 25 minutes. It is not pass/fail; it sets your baseline, and you take the parallel form at the end.
Claim, reasons, evidence and unstated assumptions; fact, value and policy; the five questions of the Claim Card; hedge words such as “up to”. Video: A slogan is not a claim. Scenario Sim: Forty per cent of what?
Three questions for any source, lateral reading, and telling an original from a repeat, so that you weigh sources by independence and not by number. Video: Down the page, or across the tabs. Scenario Sim: Four sources, one afternoon.
Anchoring, fluency, confirmation and automation bias, what the research honestly says about each, and three countermeasures: own view first, consider the opposite, a named checker. Video: The first number. Scenario Sim: We all agreed very fast.
Relative and absolute change, per cent and percentage points, plain counts, base rates behind an accuracy figure, samples, averages and chart axes. Video: One thousand orders. Scenario Sim: Redraw the deck.
Why language models go wrong, the kinds of error to expect, what to verify first by stakes and checkability, and how to verify at the original source. Video: Predicted, not looked up. Scenario Sim: Five references, forty minutes.
Assumptions under a conclusion, the other reasons two things move together, regression to the mean, and a table of rival explanations against the evidence. Video: It would have fallen anyway. Scenario Sim: Four explanations.
Framing a decision with more than two options, reversible and irreversible choices, the premortem, a go rule and a stop rule written before the data, and when to stop analysing. Video: Two doors. Scenario Sim: Before we sign.
Arguing to find out, not to win: the steelman, useful disagreement, and the one-page reasoned recommendation with the best counter-argument and what would change your mind. Video: Build their bridge first. Scenario Sim: The case, and Ines.
The final assessment, loaded from Form B: 20 parallel scenario questions, about 25 minutes. Score 70% or more (3 attempts) for the credential.
Finishing this pathway issues a credential backed by the evidence you produced along the way. It is verifiable by link and is also issued as an Open Badge 2.0, so an employer can check it without taking your word for it.
Module 5 explains why a language model can be fluent and wrong, what to verify first, and how to verify it outside the tool. The sim Five references, forty minutes puts you in front of a briefing before a meeting.
Module 2 teaches three questions for any source and the habit of leaving the page to check it. In Four sources, one afternoon you decide what to check and when to stop.
Module 3 covers anchoring, fluency, confirmation and automation bias, with what the research supports and where it stops. The sim is We all agreed very fast.
Module 4 covers relative and absolute change, percentage points, base rates behind an accuracy figure, samples and chart axes. In Redraw the deck you restate three slides fairly.
Module 6 tests an “it worked” claim with assumptions, rival explanations and a comparison group. The sim is Four explanations.
Modules 7 and 8 cover reversible and irreversible choices, the premortem, a go rule written before the data, the steelman and the one-page recommendation, practised in Before we sign and The case, and Ines.
What comes with the course, and what the practice is like.
Forty per cent of what?, Four sources, one afternoon, We all agreed very fast, Redraw the deck, Five references, forty minutes, Four explanations, Before we sign and The case, and Ines. Each is a branching browser simulation in which you make the main character’s decisions and see what follows. No real AI tool is called, and you need no account with any AI product.
One short concept video in every module, such as A slogan is not a claim and One thousand orders. They are animated and narrated by a synthetic voice, with captions and a full transcript.
The Claim Card, the Source Map, the trap check, the Number Restater, the AI Output Check, the Rival Explanations table, the Premortem and Decision Rule sheet, and the One-Page Recommendation.
Each module ends with checked links to free material from universities, standards bodies and other original sources. They are links to those sources’ own pages, not partnerships or endorsements.
Three complimentary Pro Kits from the Bodhih store come with the course, at no extra charge.
Company seats with reporting on each learner’s progress and before-and-after assessment scores. The apply-at-work tasks in Modules 3, 7 and 8 work well done with colleagues.
Ask about team seats| What is compared | A typical critical thinking course | Critical Thinking in the AI Era at Bodhih |
|---|---|---|
| AI output | Often not covered, or covered as one short section. | Three of eight modules deal with it directly: thinking traps including automation bias, numbers, and checking AI output at the source. |
| Practice | Reading, video and a quiz; sometimes peer review, where feedback depends on who reviews. | A branching Scenario Sim in every module, which gives every learner the same feedback, and a task on your own work. |
| How it is taught | A list of fallacies or frameworks. | One case across eight chapters, with a mistake the main character actually makes in each module and its repair. |
| The evidence | Well-known findings repeated without their limits. | A register of 40 sources opened at the original on 14 November 2026. Where a famous finding could not be opened or is weaker than its reputation, the course says so. |
| Time and cost | From about five hours on a subscription platform, to about 24 hours on a university short course at about £399, to a live two-day seminar at about US$2,995 (vendor pages as read on 15 November 2026). | About 11 hours, self-paced, any time zone, at $99. |
| Honest limit | University and executive programmes offer a university’s name, and live ones offer teaching and discussion with other people. | No live classes and no human tutor. The labs are simulations, the videos use a synthetic voice, and the credential is Bodhih’s own, not accredited by an outside body. |
No ratings yet. Only people who have bought this pathway can rate it - every buyer gets a one-click rating link a couple of days after their purchase.
Straight answers on time, price, tools and the credential. If yours is not here, ask us and you will have a reply the same working day.
Any professional who has to judge claims, numbers and AI answers and pass them on to someone who will act: analysts, managers, project and operations staff, marketers, and people in HR, finance and customer roles. No technical or statistical background is needed beyond percentages.
No. You earn a Bodhih credential with an Open Badge that you can share and that others can verify. It is not accredited by a university, a regulator or any outside body, and it carries no academic credit.
No. The labs are browser simulations: branching scenarios in which you make the decisions and see the consequences. No real AI tool is called, and you need no account with any AI product. The videos are animated and narrated by a synthetic voice.
The course is $99 for an individual, and $79 per seat for companies. Until 10 January 2027 the price shown on this page is a launch price, about half the regular price. After that date the regular price applies.
About 11 hours: eight modules of 75 minutes, a baseline assessment and a final assessment. It is self-paced. We expect most people to take about 8 weeks at one module a week, and you can go faster.
Complete all eight modules and pass the final assessment at 70%. You have three attempts. There is no reviewed project.
No. It names no products in its lessons and teaches no product features. It teaches how to check what any assistant gives you. For day-to-day use of assistants, see Bodhih AI at Work; for product skills, see Bodhih Copilot and Gemini at Work.
No. The numbers module uses counts such as “3 in every 100” on purpose. A Cochrane review of 35 studies of health information found that people understand counts better than probabilities (Akl and colleagues, 2011; read 14 November 2026). The final assessment needs no calculator.
From a register of 40 sources opened at the original on 14 November 2026: peer-reviewed papers, university references, standards bodies, governments and AI builders’ own publications. Where a well-known finding could not be opened or is weaker than its reputation, the course says so.
We cannot promise that, and no honest course can. You will practise specific checks on a realistic case and on your own work. In one field study of 290 business students, a single training session made people less likely to pick the weaker, belief-confirming answer in a later case (Sellier, Scopelliti and Morewedge, Psychological Science, 2019; read 14 November 2026). That is encouraging, not a guarantee, and it was not a study of this course.
Yes. Company seats are $79 each, with reporting on each learner’s progress and before-and-after assessment scores. Each person still completes their own modules and final assessment. For seats or a rollout, write to solutions@bodhih.com or call +91 99000 11601.
The course is $99 for an individual, or $79 per seat for companies, with three Pro Kits included. You can start today with the baseline assessment and Module 1, and notice how differently the next “up to 40%” reads. For a team, write to solutions@bodhih.com or call +91 99000 11601.
Starting records where you stand today. That baseline is what the re-measurement is compared against, and it is never rewritten afterwards.
Four hours of hands-on AI literacy for every employee, with a measured before-and-after score.
Use the AI assistant in your office suite for email, documents, spreadsheets, slides and meetings, and check every draft. One switch covers both suites.
Turn business questions into answers you can trust with Excel, Power BI and AI assistants, practised in the browser-based Data Lab.
Measure your team first, plan each person’s growth after, or bring a trainer or coach in. Bodhih’s sister companies cover the rest of the journey.