IB DIPLOMA · HIGHER LEVEL

IB Math AI HL
Syllabus

Study modelling and statistical reasoning alongside matrices, networks and differential equations. This AI HL map shows the shared foundations and the additional higher-level skills.

Mathematics: Applications and Interpretation HL

WHAT THIS COURSE ASKS OF YOU

How to approach AI HL

AI HL combines substantial mathematical depth with interpretation. A statistical conclusion needs suitable data, a network algorithm needs a clearly stated objective, and a differential-equation solution needs to be checked against the model it describes.

This is an independent study guide, not the official IB syllabus document. Skills are grouped in plain language so you can plan practice; the group numbers are not IB subtopic codes. Use the official guide and your teacher’s directions for the full requirements and prior learning.

Work through the shared foundations as well as the entries marked Additional HL. An HL extension often uses several earlier ideas, so a secure explanation of a basic step still matters.

CURRENT COURSE · THROUGH NOVEMBER 2028

Your AI HL topic checklist

Choose one skill, try a question and explain the decision behind your answer.

54 study skills across five topic areas.

01

Number and algebra

29 recommended hours · 9 study skills

Use algebra, complex numbers and matrices as tools for modelling.

  • Use number forms and significant figures

    Calculate with scientific notation, round appropriately and interpret percentage error. Keep extra precision during a calculation and round the final value to suit the context.

    • Use the required significant figures or decimal places only at the final step. Rounding intermediate values can distort a small difference or a later statistical result.
    • Absolute error measures the size of a difference. Relative or percentage error compares that size with a non-zero exact or accepted value. State which reference value you use.
    Scientific notation
    a×10k(1≤∣a∣<10)a\times10^k\quad(1\leq|a|<10)
    Percentage error
    percentage error=∣approximate−exactexact∣×100%\text{percentage error}=\left|\frac{\text{approximate}-\text{exact}}{\text{exact}}\right|\times100\%
  • Model additive and multiplicative sequences

    Choose an arithmetic or geometric model from the way a quantity changes. Calculate terms or finite sums and explain when extrapolating a sequence is unreasonable.

    • Use a constant difference for additive change and a constant ratio for multiplicative change. Check that the units and the meaning of n match the context.
    • A finite sum describes the total of successive quantities, not simply the final term. If r = 1, the geometric sum is nu₁ instead of the fractional form below.
    Arithmetic sequence and sum
    un=u1+(n−1)d,Sn=n2(u1+un)u_n=u_1+(n-1)d,\qquad S_n=\frac n2(u_1+u_n)
    Geometric sequence and sum
    un=u1rn−1,Sn=u1(1−rn)1−ru_n=u_1r^{n-1},\qquad S_n=\frac{u_1(1-r^n)}{1-r}
  • Investigate savings, loans and depreciation

    Use compound growth and financial technology to analyse changing balances, annuities and amortisation. State the payment period and interest convention used in a model.

    • Match the interest rate to the payment period in a finance solver. Record present value, future value, payment, period count, and whether payments occur at the beginning or end.
    • In the balance recurrence below, i is interest per payment period and p a payment deducted at its end. A loan payment includes both interest and principal; a depreciation model usually assumes a fixed fractional loss per period.
    Compound interest, annual percentage R
    A=P(1+R100k)ktA=P\left(1+\frac{R}{100k}\right)^{kt}
    End-of-period loan payment model
    Bn+1=(1+i)Bn−pB_{n+1}=(1+i)B_n-p
    Depreciation at fractional rate d
    Vn=V0(1−d)nV_n=V_0(1-d)^n
  • Use powers, logarithms and numerical solutions

    Evaluate logarithms and exponentials and solve relevant equations with technology. Interpret an exponent or a numerical solution in the original situation and reject an invalid input.

    • A logarithm answers an exponent question and requires a positive argument and a valid base. On technology, distinguish natural logarithm from base-ten logarithm.
    • Solve a model numerically within its practical domain and substitute back to check the output. A numerical solver can return a mathematically valid value that has no contextual meaning.
    Exponential and logarithmic relationship
    ax=b⟺x=log⁡aba^x=b\quad\Longleftrightarrow\quad x=\log_a b
    Solving a growth model
    Aekt=B⟹t=ln⁡(B/A)k(A,B>0, k≠0)Ae^{kt}=B\quad\Longrightarrow\quad t=\frac{\ln(B/A)}k\quad(A,B>0,\ k\ne0)
  • Work with rational exponents and logarithmic laws

    Additional HL

    Simplify expressions, change bases and solve the required exponential or logarithmic equations. Relate an exact algebraic form to a numerical result and check the domain.

    • Use exponent and logarithm laws to reorganise a model before solving it. A fractional real power may restrict the allowable inputs.
    • For logarithms, require a positive argument and a positive base different from one. A sum inside a logarithm does not split using the product law.
    Exponent laws
    aman=am+n,(am)n=amna^ma^n=a^{m+n},\qquad(a^m)^n=a^{mn}
    Logarithm laws and change of base
    log⁡a(xy)=log⁡ax+log⁡ay,log⁡ax=ln⁡xln⁡a\log_a(xy)=\log_a x+\log_a y,\qquad\log_a x=\frac{\ln x}{\ln a}
    Power law
    log⁡a(xr)=rlog⁡ax\log_a(x^r)=r\log_a x
  • Use a convergent infinite geometric series

    Additional HL

    Check the size of the common ratio before summing to infinity. Explain how the limiting sum differs from a finite financial model and reject a divergent case.

    • For an infinite geometric series, the remaining tail tends to zero only when |r| < 1. A ratio of negative magnitude below one also converges.
    • A perpetual mathematical sum is not the same as a finite repayment or savings model. State the assumptions if using an infinite series in a context.
    Infinite geometric sum
    S∞=u11−r,∣r∣<1S_\infty=\frac{u_1}{1-r},\qquad |r|<1
  • Represent complex numbers

    Additional HL

    Use Cartesian, polar and exponential forms and interpret powers or roots in the complex plane. Keep the modulus and argument connected to their geometric meaning.

    • Cartesian form describes horizontal and vertical components; polar form gives a magnitude and direction. Select the correct quadrant for the argument.
    • Multiplication multiplies moduli and adds arguments. For a non-zero number, list all n roots using integer k from zero to n − 1; do not report only the principal root.
    Complex representations
    z=a+bi=reiθ,r=a2+b2z=a+bi=re^{i\theta},\qquad r=\sqrt{a^2+b^2}
    Powers and roots
    zn=rneinθ,wk=r1/nei(θ+2kπ)/nz^n=r^ne^{in\theta},\qquad w_k=r^{1/n}e^{i(\theta+2k\pi)/n}
  • Calculate and interpret matrices

    Additional HL

    Use matrix operations, determinants and inverses to solve or represent a system. Check dimensions and recognise when a matrix is singular.

    • Check matrix dimensions before adding or multiplying. Multiplication is generally not commutative; the order represents the order of transformations.
    • For a square matrix with non-zero determinant, an inverse can solve a compatible linear system. A singular matrix has no inverse, so analyse the system rather than forcing an inverse operation.
    Two-by-two determinant and inverse
    A=(abcd),det⁡A=ad−bcA=\begin{pmatrix}a&b\\c&d\end{pmatrix},\quad\det A=ad-bc
    Inverse when ad − bc ≠ 0
    A−1=1ad−bc(d−b−ca)A^{-1}=\frac1{ad-bc}\begin{pmatrix}d&-b\\-c&a\end{pmatrix}
    Linear system
    Ax=b⟹x=A−1bA\mathbf x=\mathbf b\quad\Longrightarrow\quad\mathbf x=A^{-1}\mathbf b
  • Use eigenvalues and eigenvectors

    Additional HL

    Find the required eigen-information and apply diagonalisation to suitable matrices. Interpret repeated transformations or long-term behaviour in the setting of the problem.

    • An eigenvector is non-zero and retains its direction under the transformation; its eigenvalue scales it. Find eigenvalues from the characteristic equation.
    • Diagonalisation needs enough independent eigenvectors. Use the same eigenvector order in P and eigenvalue order in D; this simplifies repeated transformations.
    Eigenvalue equation
    Av=λv,v≠0,det⁡(A−λI)=0A\mathbf v=\lambda\mathbf v,\quad\mathbf v\ne\mathbf0,\qquad\det(A-\lambda I)=0
    Diagonalisation and powers
    A=PDP−1,An=PDnP−1A=PDP^{-1},\qquad A^n=PD^nP^{-1}
02

Functions

42 recommended hours · 9 study skills

Compare models and representations rather than treating a fit as a final answer.

  • Interpret lines and function notation

    Connect gradients, intercepts, domain and range with a context. Distinguish an input from an output and keep their units visible when evaluating a function.

    • Interpret a gradient as output units per input unit and an intercept as the predicted output at input zero. An intercept outside the observed domain may have limited practical meaning.
    • A domain restriction can come from physical constraints or the interval used to fit a model. State the range corresponding to that domain.
    Line and gradient
    y=mx+c,m=y2−y1x2−x1y=mx+c,\qquad m=\frac{y_2-y_1}{x_2-x_1}
    Function evaluation
    f(a)=ma+cwhen f(x)=mx+cf(a)=ma+c\quad\text{when }f(x)=mx+c
  • Read and use graphs with technology

    Find intercepts, turning points and intersections and choose a useful viewing window. Explain why a graphical answer matches the requested feature rather than only giving a screen value.

    • Use labelled axes and a window that shows the relevant domain and scale. A graph appearing flat can be a window issue rather than a constant function.
    • A root is an input where output is zero; an intersection solves equality of two outputs. Read coordinates and units, and retain the precision requested.
    Zeros and intersections
    f(x)=0,f(x)=g(x)f(x)=0,\qquad f(x)=g(x)
    A turning-point candidate
    f′(x)=0f'(x)=0
  • Choose a model family

    Compare linear, quadratic, cubic, exponential, power and sinusoidal models in the situations required by the course. Use the behaviour of the data and the context to justify a choice.

    • A linear model gives constant absolute change; an exponential gives constant fractional change. A power model relates changes in scale, while a sinusoid represents repeating behaviour.
    • Use a polynomial only where its shape and domain are justified. A cubic can turn twice but need not do so; a quadratic has a single turning point.
    Polynomial models
    y=mx+c,y=ax2+bx+c,y=ax3+bx2+cx+dy=mx+c,\qquad y=ax^2+bx+c,\qquad y=ax^3+bx^2+cx+d
    Exponential and power models
    y=abx (b>0),y=axny=ab^x\ (b>0),\qquad y=ax^n
    Degree-based sinusoidal model
    y=asin⁡(bx+c)+d,T=360∘∣b∣y=a\sin(bx+c)+d,\qquad T=\frac{360^\circ}{|b|}
  • Find parameters and make predictions

    Use given information or regression technology to fit a model, then interpret its parameters. Check the practical domain and distinguish interpolation from extrapolation.

    • Use enough independent information to determine the unknown parameters. A regression curve estimates a relationship from data and generally does not pass through every observation.
    • Interpolation stays within the observed input range; extrapolation goes beyond it. Report a prediction with units and with accuracy appropriate to the measurements.
    Two-point linear fit
    m=y2−y1x2−x1,c=y1−mx1m=\frac{y_2-y_1}{x_2-x_1},\qquad c=y_1-mx_1
    Substitution into a fitted exponential
    y^=abx\widehat y=ab^x
  • Evaluate and refine a model

    Compare predictions with observations and discuss limitations or unrealistic long-term behaviour. A close fit to the available data does not make a model universally valid.

    • Check a model against observations and describe a plausible source of mismatch. A residual is an observed output minus its predicted output.
    • A pattern in errors suggests missing structure. Also check limits: negative populations or endlessly increasing capacities can reveal a model’s practical boundary.
    Residual
    ei=yi−y^ie_i=y_i-\widehat y_i
    Example exponential growth multiplier
    y^(x+1)y^(x)=b\frac{\widehat y(x+1)}{\widehat y(x)}=b
  • Extend transformations and function operations

    Additional HL

    Compose and invert the required functions and analyse transformations. Explain a restriction on an inverse and use graph features to check an algebraic result.

    • A horizontal change inside a function and a vertical change outside it have different effects. Track points as well as viewing a graph.
    • Composing functions is order-sensitive. To find an inverse, restrict a many-to-one model to an appropriate branch and check the practical domain.
    Transformation
    g(x)=af(b(x−h))+kg(x)=a f(b(x-h))+k
    Composition and inverse
    (f∘g)(x)=f(g(x)),f−1(f(x))=x(f\circ g)(x)=f(g(x)),\qquad f^{-1}(f(x))=x
  • Linearise and compare nonlinear models

    Additional HL

    Use suitable logarithmic transformations and the prescribed regression or model-comparison methods. Keep transformed variables distinct from the original measured quantities.

    • An exponential model becomes linear when the output is logged; a power model becomes linear when both variables are logged. Inputs to the logarithms must be positive.
    • A transformed slope or intercept is not automatically the original parameter. Transform back, then evaluate predictions in the original units.
    Linearising an exponential
    y=Aekx⟹ln⁡y=ln⁡A+kxy=Ae^{kx}\quad\Longrightarrow\quad\ln y=\ln A+kx
    Linearising a power
    y=Axn⟹ln⁡y=ln⁡A+nln⁡xy=Ax^n\quad\Longrightarrow\quad\ln y=\ln A+n\ln x
  • Investigate extended model families

    Additional HL

    Work with the course’s additional exponential, logistic, trigonometric and piecewise models. Interpret parameters, limiting behaviour and discontinuities in context.

    • A logistic model has a limiting capacity under the assumptions encoded in its parameters. Interpret its initial level, growth rate and eventual behaviour together.
    • For a piecewise model, state each interval and examine the join. A discontinuity may describe a real change in a tariff or policy, whereas an unexplained jump can expose a poor model.
    Logistic model
    y=L1+Ae−kx(L,A,k>0)y=\frac L{1+Ae^{-kx}}\quad(L,A,k>0)
    Radian-based periodic model
    y=asin⁡(bx+c)+d,T=2π∣b∣y=a\sin(bx+c)+d,\qquad T=\frac{2\pi}{|b|}
  • Evaluate model fit rather than a single coefficient

    Additional HL

    Use residual information and the required goodness-of-fit comparisons to judge competing models. Explain why a model can fit observed data yet produce implausible extrapolations.

    • A small aggregate error does not remove the need to inspect residuals and assumptions. Compare models on the same response variable, dataset and scale.
    • More parameters can reduce in-sample residual error while weakening interpretation or out-of-range reliability. Judge the shape and context as well as the statistic.
    Residual sum of squares
    SSE=∑i(yi−y^i)2SSE=\sum_i(y_i-\widehat y_i)^2
    Coefficient of determination, when defined
    R2=1−SSE∑i(yi−yˉ)2R^2=1-\frac{SSE}{\sum_i(y_i-\bar y)^2}
03

Geometry and trigonometry

46 recommended hours · 10 study skills

Translate movement, location and network decisions into mathematics.

  • Measure distances, areas and volumes

    Use two- and three-dimensional geometry, coordinates and the required shapes. Match the units to length, area or volume and assess whether a measurement is sensible.

    • Use coordinate differences for length and match dimensions in area and volume. Draw compound solids as simpler pieces before combining measurements.
    • For a pyramid or cone, the height is perpendicular to its base. Include only exposed faces in surface area; a cone’s curved area uses slant height.
    Distance in three dimensions
    d=(x2−x1)2+(y2−y1)2+(z2−z1)2d=\sqrt{(x_2-x_1)^2+(y_2-y_1)^2+(z_2-z_1)^2}
    Prism, pyramid and sphere volumes
    Vprism=Ah,Vpyramid=13Ah,Vsphere=43πr3V_{\rm prism}=Ah,\quad V_{\rm pyramid}=\frac13Ah,\quad V_{\rm sphere}=\frac43\pi r^3
    Cone volume and curved surface
    V=13πr2h,Acurved=πrℓV=\frac13\pi r^2h,\qquad A_{\rm curved}=\pi r\ell
    Cylinder volume and total surface
    V=πr2h,A=2πrh+2πr2V=\pi r^2h,\qquad A=2\pi rh+2\pi r^2
    Hemisphere volume and total surface
    V=23πr3,A=3πr2V=\frac23\pi r^3,\qquad A=3\pi r^2
  • Solve trigonometric measurement problems

    Use right-triangle trigonometry, sine and cosine rules and triangle area. Represent bearings, elevations and spatial situations in a diagram before calculating.

    • Use right-triangle ratios only inside a right triangle. In any triangle, pair each named side with its opposite angle before selecting a rule.
    • Bearings are measured clockwise from north. Mark parallel horizontal lines when interpreting elevation or depression, and check alternative sine-rule angles if the data allow two triangles.
    Right-triangle ratios
    sin⁡θ=opphyp,cos⁡θ=adjhyp,tan⁡θ=oppadj\sin\theta=\frac{\mathrm{opp}}{\mathrm{hyp}},\quad\cos\theta=\frac{\mathrm{adj}}{\mathrm{hyp}},\quad\tan\theta=\frac{\mathrm{opp}}{\mathrm{adj}}
    Sine rule
    asin⁡A=bsin⁡B=csin⁡C\frac a{\sin A}=\frac b{\sin B}=\frac c{\sin C}
    Cosine rule and triangle area
    c2=a2+b2−2abcos⁡C,K=12absin⁡Cc^2=a^2+b^2-2ab\cos C,\qquad K=\frac12ab\sin C
  • Connect circular measurements and periodic motion

    Work with the course’s angle conventions, arc lengths, sectors and trigonometric models. Check the calculator’s angle mode before interpreting an answer.

    • For the degree formulas here, θ is the central angle and r the radius. Use a fraction of the full circumference or disc area.
    • For a periodic model in degree mode, period is 360° divided by the magnitude of the input multiplier. Amplitude is measured from the midline, not from the horizontal axis.
    Degree-based arc length and sector area
    s=θ360∘2πr,A=θ360∘πr2s=\frac{\theta}{360^\circ}2\pi r,\qquad A=\frac{\theta}{360^\circ}\pi r^2
    Periodic model
    y=asin⁡(bx+c)+d,T=360∘∣b∣y=a\sin(bx+c)+d,\qquad T=\frac{360^\circ}{|b|}
  • Construct perpendicular bisectors

    Find a midpoint and a perpendicular gradient to form a boundary line. Link the equation to points that are equally distant from two sites.

    • A perpendicular bisector passes through the midpoint of a segment and contains points equidistant from its two endpoints.
    • For a horizontal segment its bisector is vertical, and vice versa. Avoid dividing by zero when finding a perpendicular gradient.
    Midpoint and perpendicular finite gradients
    M=(x1+x22,y1+y22),m⊥=−1m (m≠0)M=\left(\frac{x_1+x_2}2,\frac{y_1+y_2}2\right),\qquad m_\perp=-\frac1m\ (m\ne0)
    Equidistance from two sites
    (x−x1)2+(y−y1)2=(x−x2)2+(y−y2)2(x-x_1)^2+(y-y_1)^2=(x-x_2)^2+(y-y_2)^2
  • Interpret Voronoi regions

    Construct or read nearest-site regions and use their edges and vertices in location problems. Explain the assumptions made when distance is used as the deciding factor.

    • A Voronoi cell contains the points nearest to its site, under the distance measure used. Edges are portions of perpendicular bisectors, not necessarily the complete bisector lines.
    • At an ordinary vertex, three or more sites are equally close. When evaluating a remote facility, compare candidate vertices and relevant domain boundaries; straight-line distance ignores roads, terrain and capacity.
    Region belonging to site A
    d(P,A)≤d(P,B)for every other site Bd(P,A)\leq d(P,B)\quad\text{for every other site }B
  • Extend circular and trigonometric reasoning

    Additional HL

    Use radians and the required trigonometric relationships in models. Link angle, frequency and period and check whether a phase shift matches the stated starting condition.

    • Use radians consistently in calculus-based periodic models. A full cycle is 2π radians and angular frequency connects ordinary frequency to period.
    • Choose amplitude, midline and phase from the context. Verify a proposed model at time zero and after one complete period.
    Radians and circular measurement
    s=rθ,A=12r2θs=r\theta,\qquad A=\frac12r^2\theta
    Frequency, period and angular frequency
    f=1T,ω=2πf=2πTf=\frac1T,\qquad\omega=2\pi f=\frac{2\pi}T
    Pythagorean identity
    sin⁡2θ+cos⁡2θ=1\sin^2\theta+\cos^2\theta=1
  • Model movement with vectors

    Additional HL

    Represent position and displacement in the required dimensions and use scalar or vector products. Relate the calculation to a line, angle, area or motion situation.

    • Subtract position vectors for displacement and differentiate a position model for velocity. Keep a common coordinate system and time unit.
    • Use a scalar product for an angle and a vector product for a perpendicular direction or area. A vector product changes sign if the order of the inputs is reversed.
    Constant-velocity position
    r(t)=r0+tv\mathbf r(t)=\mathbf r_0+t\mathbf v
    Magnitude and scalar product
    ∣v∣=vx2+vy2+vz2,a⋅b=∣a∣∣b∣cos⁡θ|\mathbf v|=\sqrt{v_x^2+v_y^2+v_z^2},\qquad\mathbf a\cdot\mathbf b=|\mathbf a||\mathbf b|\cos\theta
    Vector product area
    K△=12∣a×b∣K_\triangle=\frac12|\mathbf a\times\mathbf b|
  • Represent a network as a graph

    Additional HL

    Identify vertices, edges, degrees, paths and cycles and use weighted or directed representations. Translate the real constraint before choosing an algorithm.

    • A vertex represents an object or location; an edge represents a connection. Define whether a weight means distance, cost, time or something else.
    • Distinguish a path, trail and cycle and whether an edge is directed. In an undirected graph, the sum of vertex degrees is twice the number of edges; a loop contributes two to degree.
    Degree sum for an undirected graph
    ∑v∈Vdeg⁡(v)=2∣E∣\sum_{v\in V}\deg(v)=2|E|
    Tree edge count
    ∣E∣=∣V∣−1for a finite tree|E|=|V|-1\quad\text{for a finite tree}
  • Apply network algorithms

    Additional HL

    Use Dijkstra’s algorithm, Kruskal’s or Prim’s algorithm, and the required route-inspection and travelling-salesperson methods. State the objective and explain the limitation of an approximate route.

    • Dijkstra finds shortest paths with non-negative weights. Prim and Kruskal build minimum spanning trees connecting all vertices without cycles; a shortest route and a spanning tree solve different problems.
    • Route inspection seeks a closed walk covering every edge and may require duplicating shortest paths between odd vertices. For travelling-salesperson tasks, distinguish a lower bound from an upper-bound tour obtained by a heuristic such as nearest neighbour.
    Path weight
    w(path)=∑e in pathw(e)w(\text{path})=\sum_{e\text{ in path}}w(e)
    Travelling-salesperson bounds
    L≤optimal tour length≤UL\leq\text{optimal tour length}\leq U
  • Use adjacency and transition representations

    Additional HL

    Represent connections or repeated movements with matrices. Distinguish counting walks in a graph from modelling probabilities in a stochastic system.

    • For an adjacency matrix A, the entry in row i, column j records the specified connection from i to j. Powers of A count walks, not necessarily simple paths.
    • A transition matrix describes probabilities rather than edge counts. State whether state vectors are rows or columns and match both the multiplication order and the sums-to-one convention.
    Counting walks of length k
    (Ak)ij=number of length-k walks from i to j(A^k)_{ij}=\text{number of length-}k\text{ walks from }i\text{ to }j
    Column-state transition convention
    xn+1=Pxn,∑iPij=1\mathbf x_{n+1}=P\mathbf x_n,\qquad\sum_i P_{ij}=1
04

Statistics and probability

52 recommended hours · 15 study skills

Distinguish a model, its assumptions and the evidence supplied by a sample.

  • Critique data collection

    Identify variable types, sampling methods and possible bias. Discuss the reliability of a conclusion before using a sample to describe a wider population.

    • A sampling frame defines who can be selected. Distinguish random selection, systematic intervals, population strata, quotas and readily available participants.
    • A larger sample does not automatically remove selection or response bias. State whether an observational study can support association, causation or neither.
    Proportional stratified sample allocation
    nh=nNhNn_h=n\frac{N_h}{N}
  • Read displays and summary statistics

    Use histograms, cumulative frequency curves, box plots and measures of centre or spread. Interpret grouped estimates, outliers and the effect of a linear change to the data.

    • A histogram with unequal class widths needs frequency density so area represents frequency. Read quartiles from cumulative frequencies or an ordered sample using the relevant convention.
    • Report the spread as well as the centre. Standard deviation uses the original units, and changing the data’s scale affects spread differently from adding a constant.
    Mean and population variance
    xˉ=∑fixi∑fi,σ2=∑fi(xi−xˉ)2∑fi\bar x=\frac{\sum f_ix_i}{\sum f_i},\qquad\sigma^2=\frac{\sum f_i(x_i-\bar x)^2}{\sum f_i}
    Outlier fences
    IQR=Q3−Q1,Q1−1.5IQR, Q3+1.5IQRIQR=Q_3-Q_1,\qquad Q_1-1.5IQR,\ Q_3+1.5IQR
    Linear change of spread
    σaX+b=∣a∣σX\sigma_{aX+b}=|a|\sigma_X
  • Analyse linear association

    Use scatter plots, Pearson’s correlation and regression for prediction. Distinguish a strong relationship from a causal explanation and avoid unsupported extrapolation.

    • Pearson’s r measures linear association, with values from −1 to 1. A scatter plot can reveal outliers or curved structure that r alone hides.
    • A fitted regression line predicts the response from the explanatory variable. It does not establish that changing one variable causes the other to change.
    Regression of y on x
    y^=mx+c,m=rσyσx\widehat y=mx+c,\qquad m=r\frac{\sigma_y}{\sigma_x}
    Centroid property
    yˉ=mxˉ+c\bar y=m\bar x+c
  • Use ranks when appropriate

    Calculate or interpret Spearman’s rank correlation and compare it with linear correlation. Explain why ranked data may answer a different question about association.

    • Spearman’s coefficient applies Pearson’s correlation to ranks and measures monotonic association. Give tied observations their mean rank.
    • The shortcut below assumes no tied ranks. With ties, rank the data correctly and use technology to correlate those ranks. Compare what rank association says with what a straight-line fit says.
    Spearman shortcut without ties
    rs=1−6∑di2n(n2−1)r_s=1-\frac{6\sum d_i^2}{n(n^2-1)}
  • Organise and calculate probabilities

    Use diagrams, tables and probability rules for combined or conditional events. State whether events are independent and check whether a probability is within its valid range.

    • Use a Venn diagram for overlapping events, a table for counts or a tree for stages. Conditional probability changes the reference group.
    • Independent is different from mutually exclusive. Multiply along a tree path and add only disjoint alternatives.
    Addition rule
    P(A∪B)=P(A)+P(B)−P(A∩B)P(A\cup B)=P(A)+P(B)-P(A\cap B)
    Conditioning and independence
    P(A∣B)=P(A∩B)P(B),P(A∩B)=P(A)P(B) if independentP(A\mid B)=\frac{P(A\cap B)}{P(B)},\qquad P(A\cap B)=P(A)P(B)\ \text{if independent}
  • Use discrete, binomial and normal models

    Calculate expected values and suitable distribution probabilities with technology. Check the assumptions of a binomial model and label the mean and spread of a normal model.

    • Check that a discrete distribution has non-negative probabilities summing to one. Its expectation is a long-run mean, not necessarily a possible observation.
    • A binomial model needs fixed n, independence and constant success probability. For a normal model use μ as centre and σ as spread, and interpret inverse values as percentiles.
    Expectation and binomial model
    E(X)=∑xxp(x),P(X=k)=(nk)pk(1−p)n−kE(X)=\sum_x xp(x),\qquad P(X=k)=\binom nk p^k(1-p)^{n-k}
    Binomial moments
    E(X)=np,Var⁡(X)=np(1−p)E(X)=np,\qquad\operatorname{Var}(X)=np(1-p)
    Normal standardisation
    X∼N(μ,σ2),Z=X−μσX\sim N(\mu,\sigma^2),\qquad Z=\frac{X-\mu}{\sigma}
  • Set up a statistical hypothesis test

    Write null and alternative hypotheses, choose a significance level and interpret a p-value. Distinguish insufficient evidence from proving a null hypothesis true.

    • State H₀ and H₁ in terms of the population or relationship being tested. Choose the significance level before inspecting the result.
    • A p-value is a probability of data at least as extreme as observed under the null model. It is not the probability that the null hypothesis is true. Give a contextual conclusion with the direction implied by H₁.
    Decision rule at significance level α
    p≤α⇒reject H0,p>α⇒do not reject H0p\leq\alpha\Rightarrow\text{reject }H_0,\qquad p>\alpha\Rightarrow\text{do not reject }H_0
  • Compare frequencies and population means

    Use the required chi-squared independence or goodness-of-fit test and the two-sample t-test in suitable situations. Check conditions and communicate the conclusion about the original variables.

    • For a chi-squared test, compare observed counts with expected counts under the null. An independence test uses a contingency table; goodness of fit compares one categorical distribution with a proposed model.
    • Use the required two-sample t-test for a difference in population means when its assumptions hold. Samples must be independent and the distributional conditions appropriate; check expected-count conditions for chi-squared and follow the question’s pooling instructions.
    Chi-squared statistic
    χ2=∑(O−E)2E\chi^2=\sum\frac{(O-E)^2}{E}
    Expected counts and independence degrees of freedom
    Eij=(row total)(column total)grand total,ν=(r−1)(c−1)E_{ij}=\frac{(\text{row total})(\text{column total})}{\text{grand total}},\quad\nu=(r-1)(c-1)
    Two-sample mean hypotheses
    H0:μ1=μ2,H1:μ1≠μ2 (two-sided)H_0:\mu_1=\mu_2,\qquad H_1:\mu_1\ne\mu_2\ \text{(two-sided)}
  • Plan a statistical investigation

    Additional HL

    Compare sampling and experimental designs, including the role of randomisation. Explain how the way data is collected can restrict a conclusion, even when the later calculations are correct.

    • Random sampling helps generalise to a population; random assignment helps compare treatments. Neither should be claimed when the study design did not use it.
    • Distinguish a control, blocking and matched comparisons where appropriate. Identify a potential confounder and explain how it could produce an alternative explanation.
    Sample proportion
    p^=number with the characteristicn\widehat p=\frac{\text{number with the characteristic}}n
  • Transform and combine random variables

    Additional HL

    Find expected values and variances for the prescribed linear transformations and combinations. State when independence is needed and distinguish a standard deviation from a variance.

    • Expected values add even when variables are dependent. Variances add for independent sums, while covariance matters when dependence is present.
    • For independent differences, variances still add: the negative coefficient is squared. Convert the final variance to a standard deviation only after combining it.
    Linear expectation
    E(aX+bY+c)=aE(X)+bE(Y)+cE(aX+bY+c)=aE(X)+bE(Y)+c
    Variance for independent X and Y
    Var⁡(aX+bY)=a2Var⁡(X)+b2Var⁡(Y)\operatorname{Var}(aX+bY)=a^2\operatorname{Var}(X)+b^2\operatorname{Var}(Y)
  • Use Poisson and normal approximations

    Additional HL

    Match an event-count model to its assumptions and apply suitable approximations or continuity corrections. Check whether the conditions justify the approximation.

    • A Poisson model counts independent events at a constant average rate over a specified exposure interval. Scale λ when the length of that interval changes.
    • A normal approximation to discrete counts requires suitable parameters and a continuity correction. A Poisson approximation to binomial requires large n and small p with np appropriate to the setting.
    Poisson probability and moments
    P(X=k)=e−λλkk!,E(X)=Var⁡(X)=λP(X=k)=e^{-\lambda}\frac{\lambda^k}{k!},\qquad E(X)=\operatorname{Var}(X)=\lambda
    Binomial-to-normal approximation
    B(n,p)≈N(np,np(1−p))B(n,p)\approx N(np,np(1-p))
    Continuity correction example
    P(X≤k)≈P(Y<k+0.5)P(X\leq k)\approx P(Y<k+0.5)
  • Analyse samples and confidence intervals

    Additional HL

    Use sampling distributions and the central limit theorem in the required settings. Distinguish variation in individual observations from variation in a sample mean.

    • The sample mean varies less than individual observations. For independent observations its standard error is σ divided by the square root of sample size.
    • Under the appropriate sampling conditions the central limit theorem supports an approximate normal model for a mean. A confidence level describes the long-run coverage of an interval procedure, not a posterior probability assigned to a fixed parameter.
    Sampling distribution moments
    E(X‾)=μ,Var⁡(X‾)=σ2nE(\overline X)=\mu,\qquad\operatorname{Var}(\overline X)=\frac{\sigma^2}{n}
    Mean interval with known σ and suitable normal model
    xˉ±z1−α/2σn\bar x\pm z_{1-\alpha/2}\frac{\sigma}{\sqrt n}
    Unknown σ, suitable t model
    xˉ±tn−1,1−α/2sn\bar x\pm t_{n-1,1-\alpha/2}\frac{s}{\sqrt n}
  • Extend hypothesis testing

    Additional HL

    Select the prescribed tests for means, proportions or association in the appropriate setting. Distinguish one- and two-tailed questions and interpret significance without claiming certainty.

    • Match the test to the population claim: a normal mean, Poisson mean, proportion or correlation requires different assumptions and a different null model.
    • Choose a one-sided alternative only when justified before seeing the data. Translate the result into evidence for or against the claim, and distinguish statistical significance from practical importance.
    Mean test with known σ
    z=xˉ−μ0σ/nz=\frac{\bar x-\mu_0}{\sigma/\sqrt n}
    One-proportion normal test, when justified
    z=p^−p0p0(1−p0)/nz=\frac{\widehat p-p_0}{\sqrt{p_0(1-p_0)/n}}
    Correlation hypotheses
    H0:ρ=0,H1:ρ≠0 (two-sided)H_0:\rho=0,\qquad H_1:\rho\ne0\ \text{(two-sided)}
  • Evaluate errors and test assumptions

    Additional HL

    Explain Type I and Type II errors in context and consider whether a sample and distribution meet the method’s requirements. A numerical p-value cannot repair unsuitable data.

    • A Type I error rejects a true null; its probability is the significance level of the test. A Type II error does not reject a false null and depends on the specified alternative.
    • Power is one minus the Type II error probability. Sample size, variability, effect size and significance level influence the ability to detect a difference.
    Type I, Type II and power
    α=P(reject H0∣H0 true)\alpha=P(\text{reject }H_0\mid H_0\text{ true})
    Power at a specified alternative
    β=P(do not reject H0∣H1 true),power=1−β\beta=P(\text{do not reject }H_0\mid H_1\text{ true}),\qquad\text{power}=1-\beta
  • Model a Markov process

    Additional HL

    Build and use a transition matrix, calculate later states and investigate a steady state where it exists. State the memory and transition assumptions of the model.

    • A Markov model makes the next-state distribution depend on the current state, not the earlier history. The transition probabilities must suit the time step being modelled.
    • With column-state vectors the columns of P sum to one. A stationary distribution solves Pπ = π and has total probability one; convergence to it requires suitable properties of the chain and is not automatic.
    Later state
    xn=Pnx0\mathbf x_n=P^n\mathbf x_0
    Stationary probability vector
    Pπ=π,∑iπi=1,πi≥0P\boldsymbol\pi=\boldsymbol\pi,\qquad\sum_i\pi_i=1,\quad\pi_i\geq0
05

Calculus

41 recommended hours · 11 study skills

Use analytical and numerical approaches to investigate interacting quantities.

  • Interpret a changing quantity

    Connect average and instantaneous change with a graph or context. Use a derivative to describe increasing or decreasing behaviour and state the units of the rate.

    • The average rate uses a change over an interval; an instantaneous rate is the tangent gradient at one input. Include output units per input unit.
    • Use the derivative’s sign to describe increasing or decreasing behaviour on a stated interval. A stationary point is a candidate for a local maximum or minimum, not a classification by itself.
    Average and instantaneous change
    average rate=f(b)−f(a)b−a,instantaneous rate=f′(x)\text{average rate}=\frac{f(b)-f(a)}{b-a},\qquad\text{instantaneous rate}=f'(x)
  • Differentiate polynomial models

    Use the required power derivatives and evaluate a gradient. Find tangent or normal equations and link the algebra to a sketch of the original relationship.

    • Differentiate each polynomial term using the power rule; the derivative of a constant is zero. Evaluate the derivative at the point’s input after differentiating.
    • Use the original function to find the point on the graph. A normal has negative-reciprocal gradient when the tangent gradient is non-zero.
    Power rule
    ddx(axn)=anxn−1\frac{d}{dx}(ax^n)=anx^{n-1}
    Tangent and normal
    y−f(a)=f′(a)(x−a),mnormal=−1f′(a)y-f(a)=f'(a)(x-a),\qquad m_{\rm normal}=-\frac1{f'(a)}
  • Locate stationary points and optimise

    Find candidate extrema and judge them in the model’s allowed domain. Give a contextual answer with units rather than stopping at the stationary input.

    • Form a single-variable objective using the problem’s constraints. Solve for stationary candidates and check their meaning and the permitted domain.
    • Compare endpoint values as well as stationary values for a closed interval. A mathematical minimum outside the feasible region does not solve the practical problem.
    Stationary candidates
    f′(x)=0f'(x)=0
    Sign-change classification
    + → −: local maximum,− → +: local minimum+\ \to\ -:\text{ local maximum},\qquad -\ \to\ +:\text{ local minimum}
  • Interpret integration and accumulated change

    Find the required polynomial antiderivatives and use a boundary condition. Evaluate definite integrals with technology and explain the quantity represented by an area.

    • An indefinite integral includes a constant; a definite integral measures signed accumulation. Use a supplied initial value to determine the constant when needed.
    • The trapezoidal rule approximates area using equally spaced ordinates. For geometric area, split intervals where the graph changes sign instead of allowing positive and negative areas to cancel.
    Power integration, n ≠ −1
    ∫axn dx=an+1xn+1+C\int ax^n\,dx=\frac{a}{n+1}x^{n+1}+C
    Definite integral
    ∫abf(x) dx=F(b)−F(a)\int_a^b f(x)\,dx=F(b)-F(a)
  • Approximate an area numerically

    Apply the trapezoidal rule using appropriate intervals and endpoint values. State that the result is an approximation and consider what the graph suggests about its reliability.

    • Use equally spaced inputs and calculate each ordinate before substituting into the rule. The endpoints occur once and interior ordinates twice.
    • A trapezium joins consecutive data points by a straight line. For a curve concave upward the rule overestimates its signed integral, and for a curve concave downward it underestimates; check whether the concavity changes across the interval.
    Trapezoidal rule
    ∫abf(x) dx≈h2[y0+2∑i=1n−1yi+yn]\int_a^b f(x)\,dx\approx\frac h2\left[y_0+2\sum_{i=1}^{n-1}y_i+y_n\right]
    Step size and ordinates
    h=b−an,yi=f(a+ih)h=\frac{b-a}{n},\qquad y_i=f(a+ih)
  • Differentiate further model functions

    Additional HL

    Use the prescribed exponential, logarithmic and trigonometric derivatives and the appropriate differentiation rules. Connect a second derivative with curvature and the behaviour of a model.

    • Use radians for trigonometric derivatives. The derivative of a composite includes the derivative of its inner function.
    • Use the second derivative to study concavity and how a rate is changing. Product and quotient rules are needed when factors or denominators both depend on the input.
    Elementary derivatives
    (ekx)′=kekx,(ln⁡x)′=1x,(sin⁡x)′=cos⁡x,(cos⁡x)′=−sin⁡x(e^{kx})'=ke^{kx},\quad(\ln x)'=\frac1x,\quad(\sin x)'=\cos x,\quad(\cos x)'=-\sin x
    Product and quotient rules
    (uv)′=u′v+uv′,(uv)′=u′v−uv′v2(uv)'=u'v+uv',\qquad\left(\frac uv\right)'=\frac{u'v-uv'}{v^2}
    Chain rule
    ddxf(g(x))=f′(g(x))g′(x)\frac{d}{dx}f(g(x))=f'(g(x))g'(x)
  • Extend integration to areas and volumes

    Additional HL

    Integrate the required functions, choose suitable bounds and form the required volume-of-revolution setup. Check the sign, units and region before using technology.

    • Choose an antiderivative or numerical integration according to the task. Keep the constant for an indefinite integral and use upper minus lower for a definite one.
    • To rotate a region about the x-axis, square its radius to form the disc area. A region between two radial boundaries uses the difference of squared radii.
    Useful antiderivatives
    ∫ekx dx=ekxk+C,∫1x dx=ln⁡∣x∣+C\int e^{kx}\,dx=\frac{e^{kx}}k+C,\quad\int\frac1x\,dx=\ln|x|+C
    Trigonometric integrals
    ∫cos⁡x dx=sin⁡x+C,∫sin⁡x dx=−cos⁡x+C\int\cos x\,dx=\sin x+C,\qquad\int\sin x\,dx=-\cos x+C
    Revolution about the x-axis
    V=π∫ab(R2−r2) dxV=\pi\int_a^b(R^2-r^2)\,dx
  • Analyse motion using calculus

    Additional HL

    Connect position, velocity and acceleration and distinguish speed from signed velocity. Account for direction changes when calculating total distance.

    • Position, velocity and acceleration are linked by differentiation in time. Speed is the magnitude of velocity and can never be negative.
    • Split a time interval at a direction change when adding distance. For vector motion, speed is the magnitude of the velocity vector, not the sum of its components.
    Motion derivatives
    v=drdt,a=dvdt\mathbf v=\frac{d\mathbf r}{dt},\qquad\mathbf a=\frac{d\mathbf v}{dt}
    Distance from speed
    D=∫t0t1∣v(t)∣ dtD=\int_{t_0}^{t_1}|\mathbf v(t)|\,dt
  • Build a differential-equation model

    Additional HL

    Translate a contextual rate into an equation, solve the required separable forms and use an initial condition. Explain the solution in terms of the original variables.

    • Define the dependent variable, time unit and meaning of each parameter before writing a rate equation. Growth proportional to population gives an exponential model.
    • For a separable equation integrate each side in its own variable, retain equilibrium solutions, then use an initial condition. A logistic equation adds a capacity limit.
    Exponential rate model
    dydt=ky⟹y=Cekt\frac{dy}{dt}=ky\quad\Longrightarrow\quad y=Ce^{kt}
    Separable equation
    dydt=g(t)h(y)⟹∫dyh(y)=∫g(t) dt\frac{dy}{dt}=g(t)h(y)\quad\Longrightarrow\quad\int\frac{dy}{h(y)}=\int g(t)\,dt
    Logistic rate
    dydt=ky(1−yL)\frac{dy}{dt}=ky\left(1-\frac yL\right)
  • Interpret slope fields and Euler approximations

    Additional HL

    Read the local direction of solutions and generate a numerical approximation with the chosen step. State the starting point and recognise that numerical error can accumulate.

    • A slope field shows the derivative at points rather than a single complete solution. The initial value selects a particular trajectory.
    • Euler’s method follows a tangent over a finite step. Compare smaller steps to investigate numerical stability or accumulated error; a smaller step does not remove an incorrect model or initial condition.
    Euler iteration
    tn+1=tn+h,yn+1=yn+hf(tn,yn)t_{n+1}=t_n+h,\qquad y_{n+1}=y_n+h f(t_n,y_n)
  • Study coupled systems and phase portraits

    Additional HL

    Use the prescribed systems and phase-plane representations to examine interaction and equilibrium. Relate trajectories and their behaviour to the quantities being modelled.

    • A coupled system gives the rates of two interacting quantities. An equilibrium makes both rates zero at the same point.
    • For a linear system, eigenvalues help describe local growth, decay, oscillation or saddle behaviour. A phase portrait has the state variables as axes; time is represented by movement along trajectories.
    Linear coupled system
    ddt(xy)=A(xy)\frac{d}{dt}\begin{pmatrix}x\\y\end{pmatrix}=A\begin{pmatrix}x\\y\end{pmatrix}
    Equilibrium conditions
    dxdt=0,dydt=0\frac{dx}{dt}=0,\qquad\frac{dy}{dt}=0
    Euler step for a coupled system
    xn+1=xn+hAxn\mathbf x_{n+1}=\mathbf x_n+hA\mathbf x_n

The topic hours total 210. A further 30 hours support inquiry, problem solving, modelling and the exploration, giving 240 hours overall. This is recommended course time, not a revision timetable or an exam-weighting table.

KNOW THE TASK BEFORE PRACTISING

AI HL assessment and calculator rules

These figures describe the current course. The written papers contribute 80% and the mathematical exploration contributes 20%. Check the 2029 changes if you are preparing for the revised course.

Current IB Math AI HL written assessment
ComponentTimeMarksWeightCalculator / task
Paper 1120 minutes11030%Required
Short responses
Paper 2120 minutes11030%Required
Extended responses
Paper 375 minutes5520%Required
Two extended problem-solving questions
Exploration / IACoursework2020%Your own mathematical investigation

Paper 3 timing: use 75 minutes for the current course. Some older subject briefs show 60 minutes; IB’s published course update identifies 75 minutes as the previous/current timing and 60 minutes for the revised first-assessment-2029 course. Check the timetable for your own session.

Practise with the graphing calculator you will use in assessment. Select the correct statistical procedure, graph settings and units, then explain the result in the original context. Use the permitted-model and exam-mode instructions supplied by your school. A formula booklet is available in the written examinations, but recognising a formula does not replace selecting and justifying a method.

TRY A PIECE OF THE MATHEMATICS

Keep a transition matrix and state convention consistent

For column-state vectors, use P = [[0.8, 0.3], [0.2, 0.7]]. If 100 people start in state A and none in B, find the next two states.

Show the worked explanation
  1. Each column sums to one. With x₀ = [100, 0]ᵀ, multiplication gives x₁ = Px₀ = [80, 20]ᵀ.
  2. A second step gives x₂ = P[80, 20]ᵀ = [70, 30]ᵀ. The total remains 100 because the model conserves the population.
  3. For a steady state [a, b]ᵀ with a + b = 100, the first row gives a = 0.8a + 0.3b. Hence a = 60 and b = 40.

Check the reasoning: Row-state and column-state conventions use different multiplication orders. Check the convention and the probability sums before entering a matrix into technology.

Try a fresh version

Start instead from [0, 100]ᵀ. The first state is [30, 70]ᵀ. Check the multiplication and explain the assumed transition probabilities.

Original OnlyPrepy practice example. It is not an official IB examination question or mark scheme.

REVIEW THE DECISIONS BEHIND YOUR ANSWERS

Where AI HL Students Lose Marks

AI HL combines substantial computation with decisions about models and evidence. A plausible output can conceal an inconsistent convention or an unsuitable assumption, so make those choices explicit.

Mixing matrix and state conventions

A column-state transition matrix can be used as though its states were row vectors, or the eigenvalue order can fail to match the eigenvectors in a diagonalisation.

What to practise: State the vector convention, check the appropriate row or column probability sums and verify one direct multiplication before calculating repeated states.

Selecting a network algorithm by name alone

A shortest path, minimum spanning tree, route-inspection walk and travelling-salesperson tour solve different objectives.

What to practise: Write what must be visited or connected, whether a return is required and what the weights mean. State whether the result is an optimum, a bound or a heuristic route.

Confusing observations with sample means

A normal model for individuals can be used for a sample mean without adjusting the spread, or a test can be selected without checking its assumptions.

What to practise: Identify the random quantity first. Distinguish standard deviation from standard error and justify the distribution, sample conditions and alternative hypothesis.

Treating numerical solutions as exact or universal

An Euler approximation may use the wrong current gradient or be interpreted without its step size. An equilibrium may be claimed to attract every trajectory without analysis.

What to practise: Show the initial state and one iteration, identify the time step and compare the numerical behaviour with the differential equation or phase portrait.

Reporting an advanced calculation without evaluating the model

A matrix power, fitted curve or differential-equation solution can be correct while its assumptions do not match the situation.

What to practise: Connect the result to the original variables, test a limiting or special case and explain a concrete limitation. Revisit the interpretation when later parts add information.

When you review a question, label the first error as a concept, method, accuracy or interpretation issue. Correct that step, then attempt a fresh question before reading its solution.

MAKE STRONGER ANSWERS REPEATABLE

What Separates a 5 from a 7 in AI HL

Working towards a 7 in AI HL means combining mathematical depth with reliable interpretation: select a model, manage its representations and evaluate what its result says about the original problem.

These are practice targets, not official grade boundaries. Your final grade depends on the assessed work and the boundaries for your examination session.

Developing stronger AI HL responses
SkillA habit to improveA stronger response
Matrices and repeated processesPerform the requested multiplication or power.Keep the state convention consistent, check conserved totals and explain the assumptions behind long-term behaviour.
NetworksApply a remembered algorithm to a graph.Match the algorithm to the objective and distinguish an exact solution from a tour estimate or bound.
Statistical inferenceObtain a test output or confidence interval.Identify the population quantity, justify the sampling model and communicate uncertainty with the correct interpretation.
Calculus and systemsGenerate a numerical or symbolic solution.Connect rates, initial conditions and equilibrium behaviour, and examine whether an approximation is reliable in context.
Paper 3 reasoningComplete individual calculations without connecting the parts.Use earlier results to investigate a new feature, revise assumptions when appropriate and explain the final decision.

Check your independence

Choose a transition, network or differential-equation problem with an unfamiliar setting. Can you justify the model and convention before calculating, then explain both its result and one important limitation?

FEEDBACK YOU CAN USE IN YOUR NEXT ATTEMPT

How Tutoring Helps in AI HL

AI HL tutoring can help you choose among advanced tools and explain why a result is meaningful. The aim of a session is a method you can transfer to a new problem, with conventions and assumptions you can defend.

  1. Audit the model and representation

    Begin with your definition of the state, random variable or network objective. Check the assumptions and conventions before examining the calculator steps.

  2. Connect computation to explanation

    Compare a direct matrix step with a repeated transformation, or a numerical trajectory with a slope field. Explain what remains fixed and what changes.

  3. Practise extended decision making

    Work through a Paper 3-style investigation, using earlier results and evaluating how new information changes the model or conclusion.

  4. Check independence on follow-up work

    Attempt a fresh statistical, network or systems task after feedback. Review the choice of method, calculation accuracy and the quality of your contextual evaluation separately.

What to bring to a session

Bring a matrix, network or systems problem and your written interpretation. A focused session can test one calculation by hand, examine the modelling choices and set a new task that uses the same reasoning in a different context.

For exploration support, bring your own question, reasoning and draft working. Guidance should help you make and explain your own mathematical decisions; your assessed submission must remain your own.

Explore IB tutoring or discuss AI HL support. Include your course, exam session and the specific skill or question you want to work on.

INVESTIGATE A QUESTION OF YOUR OWN

The mathematical exploration in AI HL

The internal assessment, often called the IA, is an individual mathematical exploration worth 20%. Choose a focused question that lets you explain mathematics you understand. A manageable investigation with thoughtful decisions is more useful than a collection of advanced calculations you cannot justify.

For the current course, the five criteria total 20 marks. The Use of mathematics criterion has level-specific expectations; using an impressive technique without understanding it does not establish that those expectations are met.

  • Presentation4 marks
  • Mathematical communication4 marks
  • Personal engagement3 marks
  • Reflection3 marks
  • Use of mathematics6 marks

Make the investigation yours

Explain where data or a model comes from, why a method is suitable and which assumptions affect the conclusion. Technology should support the investigation while you interpret and evaluate its output. Keep sources and intermediate work so the development of your reasoning is clear.

Follow your school’s current academic-integrity policy. Guidance can help you understand requirements, plan a method or reflect on your reasoning; a tutor must not produce or rewrite your assessed submission. The revised 2029 course uses different criteria, outlined below.

FROM A TOPIC LIST TO A NEXT STEP

Build a practical AI HL revision routine

  1. Practise matrices, distributions and numerical methods with a written record of the convention and assumptions used.
  2. Connect a model’s algebra to its long-term behaviour, constraints and limitations; do not stop at a technology output.
  3. Use extended Paper 3-style problems to practise linking results from earlier parts and revising an interpretation when new information is supplied.

Use a short review cycle

Choose one checklist skill and attempt a fresh question without its solution. Keep your original working, identify the first point where your reasoning breaks down and practise that step. Return to the idea in another question after a gap, then combine it with related topics.

Use a weekly study plan to make room for this cycle, and an exam-error log to record the decision you want to improve. Practise with the IB Math AI HL May 2026 Paper 1 and 2 solved step by step. If you need guided practice, explore IB tutoring and include your exact course, level and examination session.

FIRST TEACHING AUGUST 2027 · FIRST ASSESSMENT MAY 2029

What changes for AI HL in 2029?

The current checklist above must not be treated as the complete revised syllabus. IB has published the following changes and a revised subject brief. Students taking the current course through November 2028 should continue to prepare its required content.

  • The published removals include complex numbers, the vector product and the Poisson distribution.
  • The stated test removals cover a normal-population mean, a Poisson-population mean and a correlation coefficient. Consult the new guide for the precise retained scope.
  • Paper 1 and Paper 2 each reduce from 110 to 100 marks. Paper 3 reduces from 55 to 50 marks and from 75 to 60 minutes; the exploration remains 20% with revised criteria.
Recommended topic hours: current and first-assessment-2029 AI HL
Study areaCurrentFrom 2029
Number and algebra2927
Functions4245
Geometry and trigonometry4645
Statistics and probability5252
Calculus4141
Inquiry and exploration3030

IB reorganises the revised content under its updated topic labels; the table groups comparable areas for orientation. Hours are teaching recommendations. They do not specify the percentage of exam marks.

Revised exploration criteria

From first assessment 2029, the exploration uses four criteria at both levels: Problem specification (4), Abstraction (6), Computation (4) and Interpretation (6). Use the appropriate criteria for your own session rather than combining them with the current five-criterion scheme.

Confirm the revised course requirements and assessment criteria with your teacher before planning for your new examination session.

CHOOSE THE MATHEMATICS YOU WANT TO STUDY

AI HL vs the other IB math courses

Start with your interests, current foundations and future course requirements. AA and AI study mathematics with different emphases. HL adds substantial breadth and depth; calculator access does not make AI an easier version of AA.

All four courses at a glance · current course through November 2028
CourseMathematical focusTime and assessmentWho might prefer it?
AA SLAlgebraic reasoning, functions and calculus with exact methods.150 hours · 2 written papers
Paper 1 without a calculator; calculator required for remaining papers.
Exploration: 20%.
You enjoy understanding how an algebraic method works and want a standard-level course with substantial symbolic practice.
AA HLDeeper proof, complex numbers, vectors and calculus alongside the shared AA foundations.240 hours · 3 written papers
Paper 1 without a calculator; calculator required for remaining papers.
Exploration: 20%.
You enjoy abstract problems and sustained mathematical arguments, and are ready to give a higher-level course regular practice time.
AI SLData, practical models, statistical interpretation and technology-supported decisions.150 hours · 2 written papers
Calculator required for every written paper.
Exploration: 20%.
You prefer interpreting mathematics in context and want a standard-level course built around modelling and data.
AI HLYou are hereAdvanced modelling and statistics, matrices, networks and differential equations.240 hours · 3 written papers
Calculator required for every written paper.
Exploration: 20%.
You enjoy data-rich problems and using technology to investigate substantial mathematics, and can commit to higher-level depth.

Is AI HL a good fit for you?

You enjoy data-rich problems and using technology to investigate substantial mathematics, and can commit to higher-level depth.

Useful starting strengths: Secure algebra and functions, careful statistical reasoning and fluency with graphing and matrix technology. You can develop these skills through practice; discuss your current readiness and the full Diploma workload with your teacher.

AI HL or AI SL?

Consider AI SL if contextual modelling appeals to you but a standard-level course fits your workload and meets your future requirements. It omits the additional HL matrices, networks and advanced statistical and calculus work.

See the AI SL topics and paper rules →

AI HL or AA HL?

Consider AA HL if exact symbolic methods, proof and advanced analytical calculus are your main interests. Both offer substantial mathematics, while AA has a non-calculator paper and a different set of extensions.

See the AA HL topics and paper rules →

AI HL or AA SL?

Consider AA SL if you prefer the analytical approach at standard level and it meets your intended entry requirements. A switch from AI HL changes both the depth and the style of the course.

See the AA SL topics and paper rules →

Make the choice in this order

  1. Check your intended destinations. Read the exact mathematics subject, level and grade required for each university programme and entry year. Do this before choosing a course on the basis of a career label.
  2. Try the actual style. Work through sample algebra and reasoning problems alongside modelling and data questions. Notice which explanations interest you and where your foundations need work.
  3. Plan the workload. Discuss HL subjects, school availability, support and study time with your teacher or Diploma coordinator. HL recommends 90 more hours than SL, alongside deeper content.
  4. Confirm the cohort. For assessment in 2029 or later, use the revised guide. Course choice and university recognition need to be checked against the requirements for your own year.

Neither an interest in engineering nor an interest in business selects a mathematics course by itself. Requirements vary by institution and programme; this comparison does not promise admission or equivalence.

A LITTLE MORE CLARITY

AI HL syllabus questions

Check your course, exam session and next practice step.

Tell us about your needs
What does the IB Math AI HL syllabus cover?

IB Math AI HL studies number and algebra, functions, geometry and trigonometry, statistics and probability, and calculus, alongside the mathematical exploration. HL includes the shared foundations and additional higher-level content; the checklist marks the additional groups.

Can I use a calculator in AI HL?

A graphic display calculator is required for all written AI papers. You still need to show the mathematical setup and interpret the output; technology does not replace a justified explanation.

How many written papers are there in AI HL?

There are two 120-minute papers worth 30% each and a problem-solving Paper 3 worth 20%, plus the exploration worth 20%. Current Paper 3 is 75 minutes; the first-assessment-2029 version returns to 60 minutes.

Is this the AI HL syllabus for 2029?

The detailed checklist covers the current course, whose final assessment is November 2028. The separate 2029 section summarises published changes, not the complete revised guide. Match the syllabus to your assessment session and confirm it with your teacher.

How is AI HL different from AA HL?

AI emphasises modelling, data, statistical reasoning and interpretation with technology. AA gives particular emphasis to algebraic and analytical reasoning and includes a non-calculator paper. Neither course should be chosen solely from a label such as easy or hard.

Are the topic hours the percentages of marks on the exam?

No. They are recommended teaching hours, not a guaranteed distribution of exam marks. A question can link several topics. Keep a balanced practice plan instead of using the hours to predict which questions will appear.

Does checking a skill mean I have mastered it?

No. A checkmark records that you reviewed a skill on this browser. Test your understanding with a fresh question, explain your reasoning and revisit it after a gap. The checklist is not a score prediction.

Can I get guidance with my mathematical exploration?

You can ask for help understanding the criteria, planning a method and reflecting on your own reasoning. Follow your school’s academic-integrity rules. The investigation and submitted writing must remain your own.