Why the IB Math AI exploration decides the 7 before the exam papers do

IB Math: Applications & Interpretation is the modelling-led route through the IB Diploma mathematics group, built for students who want technology to carry the algebraic load so the thinking can stay on the problem. Across both HL and SL the course shares a common spine: number and algebra, functions, geometry and trigonometry, statistics and probability, calculus, and—for HL—a deeper statistics and functions block. What makes AI distinctive is not the topic list alone; it is the way the assessment rewards students who can choose a tool, set it up correctly, and interpret the output. Every Paper 2 question is built on the assumption that a graphic display calculator is in the candidate's hand, and the mark scheme awards marks for what the machine cannot do—formulating the model, stating assumptions, and writing the answer in context. This article centres on that assessment reality, the specific places where candidates lose marks despite "knowing the topic", and the preparation strategy that converts syllabus knowledge into a band 7.

What the IB Math AI syllabus actually tests under exam conditions

The most common misreading of IB Math: Applications & Interpretation is that it is "easier" than Analysis & Approaches because the algebra is lighter. The marking data tells a different story. Paper 2 in particular is constructed so that raw computation is rarely the differentiator; the marks are clustered around interpretation, justification, and command-term response. For most candidates reading this, the shift in mindset is the single largest gain available before any new content is touched.

At SL the assessment consists of Paper 1 (short-response, calculator-compulsory, 90 minutes, 80 marks) and Paper 2 (extended-response, calculator-compulsory, 90 minutes, 80 marks), each contributing 40 percent of the final grade, plus the internal assessment—the exploration—which contributes 20 percent. At HL, Paper 1 and Paper 2 are each 120 minutes for 110 marks, again each 40 percent of the grade, with the exploration contributing 20 percent. The exploration is internally marked by the teacher and externally moderated by the IB, and it is the only piece of work a student can revise into shape over many weeks. In practice, the exploration is where the band 5 / band 7 boundary is often decided, not because the exam papers are easy but because the exploration is forgiving of exam-room nerves.

The subject guide is explicit that AI is a modelling-led course. Topics such as exponential models, normal distribution applications, kinematics, financial mathematics, and probability trees are not taught as abstract machinery; they are taught as responses to a stated context. A candidate who can solve a quadratic but cannot decide which variable to plot is not ready for Paper 2. The skill the syllabus is really testing is the loop: read a context, choose a representation, execute on the GDC, interpret the screen, then write one sentence that says what the answer means in the world the question came from. The rest of this article is built around that loop.

Paper 2: the GDC mark scheme rewards process, not keystrokes

Paper 2 in IB Math AI is the paper most candidates underestimate on first reading. The questions are long, the prompts are friendly, and the marks look plentiful. The trap is that the marks are clustered in places the calculator cannot reach. A typical 16-mark question will give one mark for a correctly entered equation, two or three marks for an answer with units in the correct form, and the remaining marks for interpretation lines, reasonableness checks, or comments on the model's limitations. The candidate who answers in three lines has usually earned only three marks.

The verb taxonomy matters more in AI than in many other IB subjects because the command terms are graded, not just descriptive. State requires a single value or short phrase. Calculate requires a working line. Explain requires a reason tied to a feature of the data or the function. Discuss requires at least two distinct points. Evaluate requires a judgement with a numerical basis. A candidate who answers an evaluate prompt with a single rounded number has, in my experience, lost between 30 and 50 percent of the available marks for that part. The fix is mechanical: before writing, read the verb, count the mark allocation in the candidate's mental ledger, and then commit to writing that many sentences.

For most candidates the single highest-leverage revision move is to print three or four past Paper 2 scripts and annotate them strictly by command term and by mark. A useful discipline is to mark each line of a model answer with one of four labels: M (machine work, calculator), P (process, written setup), I (interpretation, sentence tying the number back to context), and C (commentary, a remark on assumption, limitation, or reasonableness). Roughly 60 percent of marks in AI Paper 2 sit in the P and I categories. Candidates who only practise M lines will plateau at a band 4 or low band 5 regardless of how fluent their keystrokes are. IB Courses' one-to-one IB Math AI programme walks each student through a Paper 2 deconstruction of this kind, line by line, until the M / P / I / C reflex is automatic.

Worked example: a Paper 2 normal-distribution item

Consider an item in the style of, but distinct from, a real paper: "The masses of packets of flour produced on a production line are normally distributed with mean 502 g and standard deviation 1.6 g. A quality inspector rejects any packet that is more than two standard deviations from the mean. Estimate the probability that a randomly chosen packet is rejected." The GDC gives an answer in roughly three keystrokes. The marks, however, are distributed: one mark for the correct parameter entry, one mark for the correct shading direction in the normalcdf call, one mark for the final probability to four decimal places, and one mark for an interpretation sentence such as "Approximately 4.6 percent of packets are expected to be rejected, so the production line is producing about 4 to 5 percent waste." That final sentence is the line that moves a candidate from a band 5 to a band 6 in a mark scheme built around AI's modelling ethos.

Paper 1: the short-response trap and the minutes-per-mark budget

Paper 1 in IB Math AI is short-response only. The marks per question are lower, the prompts are sharper, and the time pressure is real. At SL the 80 marks must be cleared in 90 minutes, which is a budget of just over one minute per mark; at HL the 110 marks in 120 minutes are a little more generous, but not by much. The most common error is to spend four minutes on a two-mark part and arrive at Paper 1's later questions with no time for the geometry or probability items that are actually the easiest marks on the paper.

The short-response format also changes what the command term "calculate" requires. A two-mark calculation must show enough working to justify the answer if the final value is wrong, but it does not require an interpretation sentence. A three-mark calculation, by contrast, often does. The minute budget is best managed by reading every question once at the start of the paper, sorting them mentally into "secure" and "tentative", and answering the secure set first. This is not a speed-reading trick; it is a triage that prevents a slow first item from silently compressing the rest of the paper.

For Paper 1 specifically, the highest-yield preparation move is a fluency drill. A candidate should be able to enter any of the following into a GDC and arrive at an answer in under 90 seconds: a normal distribution probability, a regression line and its r value, a system of three linear equations, a definite integral with bounds, a future value of an annuity, and a present-value discount. Those six operations account for a large share of the short-response marks across recent AI papers. IB Courses structures its IB Math AI Paper 1 clinic around exactly this drill, with a stopwatch and a 12-item weekly set that is graded on both correctness and time.

The exploration: why the internal assessment is where the 7 is won

The exploration is the most underestimated component of IB Math: Applications & Interpretation. It contributes 20 percent of the final grade, is marked against five criteria (presentation, mathematical communication, personal engagement, reflection, and use of mathematics), and is the only piece of work a student can write, redraft, and polish over a period of weeks. A candidate who is aiming at a band 7 and who treats the exploration as a single weekend draft is leaving marks on the table that the exam papers will not return.

The five criteria are unequal in their weight in practice. Use of mathematics is the most generous: it rewards accurate, relevant, and—at HL—rigorous mathematics that sits at or above the level of the syllabus. Personal engagement is the most subtle: it is awarded for independent thinking, the choice of a non-trivial context, and the candidate's own voice. A common mistake is to read "personal engagement" as biographical, when in fact it means showing the examiner that the student is driving the mathematics, not the other way round. Reflection, introduced in the revised subject guide, is the criterion most candidates score lowest on, because reflection is not a summary; it is a statement of what the candidate would change, what they would extend, and what the limitations of the model are.

A workable exploration structure is five pages: a one-page introduction that names the context and the question, one page of raw data or function definitions, one page of method with the GDC outputs, one page of interpretation, and one closing page of reflection. That word- and page-budget is conservative; it leaves room for diagrams and screenshots and it forces the candidate to make every sentence carry weight. The exploration is also the place where the modelling ethos of AI shows up most clearly, because the candidate must choose a context that genuinely suits the toolset—finance, population, sports analytics, traffic flow—and not a context that is forced onto a piece of machinery the candidate happens to like. In my experience this is the single decision that most strongly correlates with the top two bands.

Common pitfalls and how to avoid them

• Treating the exploration as a single sitting. Schedule a draft one month before the deadline, a teacher-feedback week, a rewrite week, and a final polish week.
• Choosing a context that is too broad. A 4 000-word exploration of "climate change" will fail on focus; a 4 000-word exploration of "modelling June sea-surface temperature anomalies in the North Atlantic since 1982" has a chance.
• Quoting calculator outputs without context. Every GDC screen must be followed by a sentence that says what the screen means in the chosen context.
• Ignoring the reflection criterion. A closing paragraph that lists three limitations of the model and one specific extension is the cheapest source of marks on the rubric.
• Misreading command terms. "Discuss" is not a one-sentence prompt; "evaluate" is not a calculation; "comment on" is not optional.

Command terms and mark-scheme verbs: the single highest-leverage revision move

The IB command-term glossary is not a decorative document; it is the rubric for nearly every mark on every AI paper. The verbs are graded, and the marks are awarded against the verb's definition, not against a generic sense of "what a good answer looks like". For IB Math AI the most heavily weighted verbs are calculate, explain, describe, evaluate, discuss, and sketch. Each of these has a specific mark-scheme behaviour that candidates can reverse-engineer.

Calculate in AI permits a GDC answer; the marks are for setting up the calculation correctly and stating the answer to an appropriate degree of accuracy. Explain requires a reason; a restatement of the answer is not an explanation. Describe requires a feature of the model or the data, not a number. Evaluate requires a numerical basis and a judgement, in that order. Discuss requires at least two points, often on opposite sides of an issue. Sketch requires key features: intercepts, asymptotes, turning points, end behaviour—not a perfectly accurate drawing.

One tactical move that consistently lifts marks is to build a one-page "verb translation sheet" before the exam, where the candidate writes, in their own words, what each verb demands and what the cheapest acceptable answer looks like. This sheet is not a cheat sheet; it is a planning document. Candidates who internalise the verb definitions usually recover between five and ten marks across a Paper 2 simply by answering to the verb rather than to the topic. The same sheet, used during the exploration rewrite, will improve the personal engagement and reflection scores because the candidate is forced to be specific about what their work demonstrates.

Comparing AI to AA: how the same topic is graded differently

The two IB Diploma mathematics routes share most of their topic list but grade identical content very differently. The table below is a tutor's-eye comparison based on how mark schemes treat the same mathematical content, not a recommendation of one course over the other.

For a candidate who enjoys the algebraic side of mathematics and is comfortable without a calculator for extended reasoning, AA tends to suit. For a candidate who reasons well from a screen, prefers context-led problems, and plans to study a subject at university where modelling matters more than symbolic manipulation—economics, biology, geography, business, the social sciences—AI is usually the stronger fit. The decision is rarely about difficulty; it is about which mark scheme rewards the candidate's natural working style.

Building a term-by-term preparation plan for IB Math AI

A workable IB Math AI preparation plan is built around the assessment calendar, not around the textbook. The first term is for syllabus coverage and GDC fluency. The second term is for past-paper exposure and command-term drilling. The third term is for the exploration draft and exam-paper consolidation. The final weeks are for timed mocks and a structured error log.

GDC fluency in the first term means a candidate can produce, on demand and without notes, a regression line with r value, a normal-distribution probability, a system-of-equations solution, a definite integral, a present value, a future value, a graph with a window set, and a piecewise function plot. A short weekly drill—three items, stopwatch, self-mark against a mark scheme—is more effective than rereading chapters. The mark scheme is the textbook in AI; everything else is supporting material.

Past-paper exposure in the second term should be Paper 2 first. Paper 2 is the harder paper to revise well because the marks are spread across interpretation, and it is the paper where the M / P / I / C labelling exercise pays off. A candidate who has done four full Paper 2 papers under timed conditions, with a written self-mark against the official mark scheme, will arrive at the mock season with a much clearer sense of where the marks are. Paper 1 can be slotted in as a fluency check every two weeks; it does not need the same deep deconstruction because the marks per question are smaller.

The exploration should be drafted before the mock exam season begins, not after. A draft submitted to the subject teacher six weeks before the deadline gives time for two rounds of feedback, which is the minimum needed to lift a band 4 to a band 6 on the personal engagement and reflection criteria. The exploration is the candidate's only piece of work in the course where they can respond to feedback before submission; the exam papers offer no such second chance.

Reading AI mark schemes: the verbs that move you up a band

AI mark schemes are unusually friendly documents because they are detailed and they show the marks. The habit of reading a mark scheme after every practice question—not just checking the final answer—is the single most efficient revision habit a candidate can build. The mark scheme tells the candidate, in plain language, what a band-7 answer looks like on a given prompt, and the difference between a band 5 and a band 7 answer is rarely a new concept. It is usually one or two extra sentences that the lower-band answer did not include.

Three verb changes account for most of the band's upward movement. First, replacing state with explain in a candidate's mental preparation: many candidates answer an explain prompt with a state-style answer and lose two marks. Second, adding the word "because" and a reason to a single-line calculation: the calculation earns its mark, and the reason earns the second. Third, treating comment on as a one-sentence mini-reflection: the mark scheme awards the mark for any specific, model-aware remark, and candidates who skip the comment line leave the mark on the table.

For HL candidates the same habits apply, with one addition. HL Paper 2 contains a small number of marks that are reserved for mathematical rigour: a clear statement of an assumption, a correct unit on a final answer, a domain restriction that is made explicit. These are the marks that distinguish a band 6 from a band 7 in HL, and they are the marks that are easiest to teach once a candidate is looking for them. A useful exercise is to take any HL Paper 2 model answer and underline every assumption, unit, and domain statement. The density of those underlines is a proxy for the rigour of the paper.

Conclusion and next steps for IB Math AI candidates

IB Math: Applications & Interpretation rewards a specific kind of candidate: the one who can read a context, choose a tool, set it up correctly, and write one sentence that says what the screen means. The exam papers are not testing whether the candidate can perform algebra without a calculator; they are testing whether the candidate can use the calculator to do the algebra the syllabus expects, and whether the candidate can turn the resulting number into an answer that another human being can act on. The exploration rewards the same loop, with reflection and personal engagement layered on top. A preparation plan built around GDC fluency, command-term awareness, the M / P / I / C labelling habit, and an early exploration draft will reliably outperform a plan built around rereading the textbook.

For candidates who are aiming at the upper bands, the next step is a diagnostic on a single past Paper 2 with a strict mark-scheme self-mark, line by line, command term by command term. That diagnostic shows, in two hours, exactly where the marks are being lost and which sections of the syllabus the candidate is under-using. IB Courses' one-to-one IB Math AI programme turns that diagnostic into a term-by-term preparation plan, with a Paper 2 deconstruction clinic and a structured exploration-feedback cycle built into the timetable.

Frequently asked questions