Virginia Tech® home

Mathematical Induction

Image banner with student work

"Beyond its significance as a proof technique in mathematics, MI can provide a context to enhance students’ conception of proof"

-- Guershon Harel, 2001

Proof by mathematical induction poses a persistent challenge for students enrolled in proofs-based mathematics courses. Prior research indicates a number of related factors that contribute to the challenge, and suggests fruitful instructional approaches to support students in meeting that challenge. In particular, researchers have suggested quasi-induction as an intuitive approach to understanding the role of the inductive implication (Avital & Libeskind, 1978; Harel, 2001). However, a cognitive gap remains in transitioning to formal proof by induction. 

image depicting cognitive gap between quasi induction and formal induction

The gap includes the danger of inadvertently assuming what one is trying to prove. Quantification of the inductive implication and the inductive assumption require careful use of language, using terms that may seem ambiguous to students. This module includes tasks designed to build upon students’ conceptualizations of logical implication, utilize quasi-induction, and bridge the gap that remains.  

Identified Epistemological Obstacles

Mathematical induction is used to prove statements that are universally quantified over some infinite subset of the integers (e.g., P(n) for all positive integers, n). However, there comes a point in an inductive proof where students suppose the inductive hypothesis P(k). Students might see this as a circular argument, or they might overlook the potential danger and ostensibly assume what they are trying to prove.

To engage students in the formal method of mathematical induction, the instructor introduced the six scenerios shown in the image below. Students considered whether the given information was enough to determine the proposition, P, was true for all positive integers. 

Following their initial evaluations (as notated in the image), the class compared the first and fifth scenarios. Note that scenario 1 includes the supposition of the theorem to be proved. In contrast, scenario 5 assumes only the implication that P(k) implies P(k+1). 

Llc1 Class Task
Loading player for https://video.vt.edu/media/1_24snjg5c...

The instructor then used the class discussion, comparing scenarios 1 and 5, to launch group discussions in which students at each table discussed the logic behind scenario 5, as well as scenario 6.

Loading player for https://video.vt.edu/media/1_uvu9rzrm...

When the class gathered back together to discuss scenario 5, one student shared that the given information does not say whether P(2) is true, but rather "if P(2) is true, then P(3) is true." Note, she is making a distinction here about the truth of the proposition for a given value and the truth of the implication for that value.

The instructor then illustrated this distinction in the drawing shown in the image above. The drawing represents the truth of particular values of n (or k) with filled in circles; and the truth of an implication, from one value to the next, with an arrow.

Loading player for https://video.vt.edu/media/1_7egd34oh...

To further challenge the students on the potential conflation between the truth of the hypothesis and the truth of an implication, the instructor asked them to consider the following implication: "If for any k, k=k+1, then k+1=k+2." The video below comes from the Squre table, as they attempted to make sense of this implication (whose implied hypothesis is obviously false). 

Loading player for https://video.vt.edu/media/1_mr0r6f3n...

When proving the inductive implication, students are proving a universally quantified statement: for all k (or at least for all k up to n), P(k) should imply P(k+1). However, when making the inductive assumption, they cannot assume P(k) for all k. Rather, they are supposed to assume P(k) for some fixed but arbitrary value of k. Thus, there is need for a subtble shift in quantification that students might not appreciate.

After students had gained experience making informal inductive arguments, the instructor introduced formal proof by induction. She introduced it in two parts: (1) the base case, and (2) the inductive implication (see image below). 

In the following video, the instructor asks students how they might prove the inductive implication, which is universally quantified. When a student in the class suggested that the proof might begin by assuming P(k), the instructor pushed them to define and quantify k, as a fixed but arbitrary value of n.

Qai Class Task
Loading player for https://video.vt.edu/media/1_95z9hlwe...

In the next class, the instructor challenges students to complete a proof by induction, in their groups. So that the content of the proof would not interfere with the challenge of the new proof method, she chose a familar context. Students were to prove (by induction) that, for all positive integers, n, 3 divides 22n-1. The video below shows the Triangle group working toward a proof. Note that the students quickly agree that they should assume P(k) is true, they question whether they need to explicitly state that. Also, although they have explicitly (and universally) quantified the inductive implication, they don't quantify k in the inductive assumption.

Loading player for https://video.vt.edu/media/1_0qhdlgdq...

During whole-class discussion of the task, the instructor presses students to attend to the quantification of this new variable, k, that they have introduced. When pressed, students seem to agree that k should serve as a "fixed but arbitrary" value.

Loading player for https://video.vt.edu/media/1_rmy3akz1...

By their very nature, variables can take on various values, but they also serve various roles in proofs, independent of their value [Dawkins reference]. For example, in the epsilon-delta definition for the limit of a function, particular values of episilon do not matter so much as the role of epsilon as an independent variable from which we can derive the dependent variable, delta. Likewise, in proof by induction we might use n=k to signify a shift, not in the potential values of n or k, but in the role of k in signifying a fixed but arbitrary value of n.

Typically in introducing proof by mathematical induction, textbooks and instructors will introduce n as a univerally quantified variable. For instance, they will say "we need to prove P(n) for all n," or "for all n, P(n) implies P(n+1)." However, this instructor took a different approach, introducing n as a fixed but arbitrary value to be reached by mathematical induction. Thus, in making the inductive argument, from the base case to this fixed value of n, a new variable was needed. The video shown below illustrates the approach in a class discussion about the roles and values of variables.

Loading player for https://video.vt.edu/media/1_ezz3jrpe...

In the task shown below, the instructor presents the issue directly to the students, to work on in groups: "why do you think we use k instead of n?" The videos below show the Circle, Triangle, and Hexagon groups responding to this task.

Qrv class task
Loading player for https://video.vt.edu/media/1_29ypo79f...

The Triangle group:

Loading player for https://video.vt.edu/media/1_n1jbti5g...

The Hexagon group:

Loading player for https://video.vt.edu/media/1_c4qk6use...

In proving that a property holds for all values of x, we can take a fixed but arbitrary value of x and demonstrate that the property holds for that arbitrary value. This is called the principle of universal generalization (or PUG). The PUG is used in proofs of all kinds of universally quantified statements, and it relies on the logic that if something holds for each x, then it holds for all x. The PUG plays a particularly important role in proof by mathematical induction, within the proof of the inductive implication. 

The instructor elicited consideration of the PUG in two different ways: first in a group task, asking students to consider the general structure of "for all" statements; second in a whole-class discussion about how students might constructively prove "for all" statements. Both of these efforts seemed effective, as students suggested "for all values" meant the thing as "for any arbitrary value." 

Loading player for https://video.vt.edu/media/1_w148ym74...
Loading player for https://video.vt.edu/media/1_6cymc6ei...

Once students had an intellectual need for the PUG (as in contructively proving universally quantified statements), the instructor introduced PUG as a principle. 

Loading player for https://video.vt.edu/media/1_4zlavw6z...

Although students did see the intellectual need for PUG and seemed to accept the instructor's suggestions for using it to prove universally-quantified statements, the principle still appeared to be problematic for students at times, as illustrated in the following whole-class discussion.

Loading player for https://video.vt.edu/media/1_23g4o2x6...

What Research Says...

Prior research has identified the following challenges associated with students’ mastery of proof by mathematical induction: understanding the role of the base case (Baker, 1996; Ernst, 1984; Ron & Dreyfus, 2004; Stylianides et al., 2007), treating logical implication as an invariant relationship (Dubinsky, 1986, 1991), discerning between the truth of the conjecture and the inductive hypothesis (Movshovirz-Hadar, 1993), attending to (hidden) quantifiers (Shipman, 2016) and the proper use of related language (Ernst, 1984; Movshovitz-Hadar, 1993; Stylianides et al., 2007), and having domain-specific knowledge particular to the conjecture (Dubinsky, 1991). Many of these challenges, especially quantification, extend well beyond mathematical induction (Dawkins & Roh, 2020; Lew & Mejía-Ramos, 2019).