Engaging students with statistics through choice of real data context on homework

Although providing students with a choice of real-data contexts for homework did not significantly improve grades in an introductory statistics course, it substantially enhanced their engagement, motivation, and sense of autonomy, leading the authors to recommend incorporating varied, student-relevant contexts as a key pedagogical tool.

The Gist

Imagine you are a teacher trying to get your students to eat their vegetables. You know that vegetables (statistics) are good for them, but they often taste like "boring numbers" to the kids.

This paper is about an experiment two professors ran to see if they could make the "vegetables" taste better by letting the students pick their own flavor.

Here is the story of what they did, what they found, and what it means for teaching, explained simply.

The Big Question: Does Choice Make a Difference?

The professors, Catalina and Mine, wondered: If we let students choose the "flavor" of their homework, will they care more about it?

Usually, teachers pick a topic (like "coffee sales" or "music") and say, "Everyone, do this." But what if one student loves coffee and another thinks it's gross? What if one loves music and the other hates it?

To test this, they set up a little experiment in a college math class. They had two groups of students:

1. The "Pick Your Own Adventure" Group: These students got three different homework options every week. They could pick the one about AI cheating, gluten-free pasta, or sleep deprivation.
2. The "Surprise Me" Group: These students got the exact same math problems, but they were assigned a topic randomly. They had no choice.

They swapped these groups around every two weeks so everyone got a taste of both styles.

The Results: Grades vs. Feelings

1. The Grades (The Scoreboard)
Did the students who got to choose get better grades?

• The Answer: No.
• The Metaphor: Imagine two runners. One gets to pick which path to run on (forest or beach), and the other is told which path to run on. Both runners finished the race in about the same amount of time. The choice didn't make them run faster or score higher points.

2. The Feelings (The Heart)
Did the students enjoy the experience more?

• The Answer: Yes, absolutely!
• The Metaphor: Even though the runners finished at the same time, the one who picked the path felt like they were on a fun vacation. The one who was told what to do felt like they were on a boring commute.
  • Students who got to choose said they felt more motivated.
  • They felt more connected to the real world.
  • They felt a sense of freedom (autonomy).
  • They actually started thinking, "Hey, this is interesting!" instead of "Ugh, just another math problem."

What Did Students Actually Want?

The professors thought they knew what students would like. They were wrong.

• The "AI Cheating" topic was popular because it was current and relevant to their lives.
• The "Dog Lifespan" topic was avoided by many because it made them sad (it involved dogs dying).
• The "Teen Pregnancy" topic was avoided because it felt too heavy or sensitive.
• The "2003 Movies" topic was a hit in the end because students loved digging into old data they knew.

The Lesson: Students don't just want "real" data; they want data that feels like their life. They want to see themselves, their careers, or their hobbies in the numbers.

The Four Rules for Teachers (The Takeaway)

Based on this experiment, the authors give four simple rules for anyone teaching statistics:

1. Use Real Stories, Not Fake Ones: Don't make up numbers about "flipping coins" unless you have to. Use real data about real things (like sleep, movies, or AI). Real stories make the math feel like a detective game, not a chore.
2. Pick What They Care About: If you teach a class of future doctors, use medical data. If you teach gamers, use game stats. If you don't know what they care about, ask them. The professors guessed wrong every single time; they needed to listen to the students.
3. Serve a Variety of Dishes: You can't please everyone with one menu. Some students love sports, others love art. By offering a mix of topics, you increase the chances that everyone finds something they like at some point.
4. Let Them Choose (When You Can): Even if it's just a small choice (like picking between three homework topics), giving students a say makes them feel like partners in the class, not just robots. It gives them a tiny bit of control, which makes the whole experience more enjoyable.

The Bottom Line

Giving students a choice didn't magically make them smarter or get them better grades. But it did make the class feel less like a prison and more like a place where they could explore.

It turns out, when you let people pick their own path, they are more willing to walk it. For teachers, the effort to find different real-world stories and offer choices is worth it, not because it boosts test scores, but because it boosts engagement. It turns "doing homework" into "exploring the world."

Technical Summary

Here is a detailed technical summary of the paper "Engaging students with statistics through choice of real data context on homework."

1. Problem Statement

Statistics education literature (e.g., GAISE College Report) strongly advocates for teaching with real data within authentic contexts to help students perceive statistics as relevant and engaging. However, a significant pedagogical challenge remains: defining "relevance."

• The Dilemma: Instructors often struggle to select data contexts that resonate with a diverse student body. While some suggest using data relevant to students' daily lives, majors, or career goals, individual preferences vary widely.
• The Gap: There is a lack of empirical evidence regarding whether allowing students autonomy to choose their data context improves engagement or learning outcomes compared to instructor-assigned contexts. Furthermore, it is unclear which specific characteristics of data contexts drive student preference.

2. Methodology

The authors conducted a quasi-experiment in two sections of an introductory probability and statistics course at a large public university (Winter 2025).

• Participants: $N=65$ consenting students (33 in Section A, 32 in Section B) out of a total class of 180. The course was primarily for Computer Science majors.
• Design: A crossover design was used over a four-week intervention period.
  • Treatment Condition: Students were presented with three homework options per week. All options covered the same statistical concept (e.g., confidence intervals for proportions) but utilized different real-world data contexts.
  • Control Condition: Students were randomly assigned one of the three data contexts without choice.
  • Rotation: The sections alternated roles. For the first two weeks, Section A had choice (Treatment) and Section B had random assignment (Control). For the last two weeks, the roles were swapped.
• Data Contexts: The study utilized diverse real-world datasets, including:
  • Serious/Scientific: Sleep deprivation, electric vehicle adoption, teen pregnancies, gorilla lifespans.
  • Lighthearted/Pop Culture: 2003 movies, gluten-free pasta, home-team advantage, typing speeds.
  • Recycled: "Cheating with AI" was used in multiple weeks to test repetition effects.
• Data Collection & Analysis:
  1. Quantitative (Grades): A Generalized Estimating Equation (GEE) model was used to compare homework grades between treatment and control conditions, controlling for pre-intervention grades, student section, and week indicators.
  2. Qualitative (Surveys & Open-ended Questions): Post-intervention surveys and specific questions embedded in homework assignments asked students to explain their choices (or dislikes) regarding data contexts. Thematic analysis was performed on these responses.

3. Key Results

A. Quantitative Outcomes (Grades)

• Finding: There was no statistically significant difference in homework grades between the treatment (choice) and control (no choice) groups ($p = 0.2377$).
• Interpretation: Providing choice did not directly translate to higher scores. The authors note potential confounding factors, such as low statistical power, the "ceiling effect" (many students already scored near perfect), and the complexity of modeling within-student correlations.

B. Qualitative Outcomes (Student Sentiment & Engagement)

Thematic analysis revealed four major categories of student experience:

1. Sentiment:
   • Positive (43/65): The majority reported increased motivation, engagement, and enjoyment.
   • Impartial (18/65): Many felt choice made no difference to their learning or enjoyment.
   • Negative (2/65): A small minority felt choice added unnecessary cognitive load ("decision fatigue") or made peer discussion harder due to divergent assignments.
2. Perceived Benefits:
   • Autonomy: Students valued the "small privilege" of choice, which made assignments feel less like a chore and more like an exploration.
   • Relevance: Students reported that working with data they cared about made statistics feel "real" and applicable to the world, rather than abstract calculation.
   • Deep Learning: Choice led to deeper engagement, with students asking questions about study design, bias, and exploring variables beyond the required homework questions.
3. Preferred Context Characteristics:
   • Relevance: Students overwhelmingly preferred contexts tied to their field of study, career path, personal interests, or daily life.
   • Familiarity vs. Novelty: Preferences were complex. Some wanted familiar topics; others wanted to explore unknown outcomes.
   • Sensitivity: Topics involving death (dog lifespans) or sensitive social issues (teen pregnancy) were often avoided unless introduced carefully.
   • Instructor Miscalculation: The instructors' predictions of which topics would be most popular were consistently wrong, highlighting the difficulty of guessing student interests.

4. Key Contributions

The paper contributes to statistics education research by:

1. Empirical Evidence on Choice: Demonstrating that while choice does not necessarily boost grades, it significantly enhances subjective engagement, motivation, and perceived autonomy.
2. Taxonomy of Relevance: Moving beyond the vague concept of "relevance" to identify specific student-driven characteristics (career alignment, personal experience, curiosity about unknown outcomes) that make data contexts effective.
3. Pedagogical Framework: Providing a structured approach for instructors to move from "simulated/made-up" data to "real cleaned" data with context, and how to implement choice effectively.

5. Significance and Recommendations

Based on the findings, the authors propose four recommendations for statistics educators:

1. Use Real Data with Real Contexts: Avoid artificial data where possible. Real data provides the necessary depth for students to understand the limitations and power of statistical analysis in the real world.
2. Select Contexts Students Care About: Instructors should not rely solely on their own intuition of what is "interesting." Instead, they should survey students or use diverse datasets to find connections to student careers, majors, and daily lives.
3. Incorporate Variety: No single context appeals to everyone. A diverse set of contexts (scientific, social, pop culture) ensures that different students find entry points for engagement throughout the course.
4. Consider Choice as a Pedagogical Tool: Even if it doesn't raise grades, offering choice fosters autonomy and motivation. Instructors can mitigate the grading burden by keeping the statistical tasks identical while varying the data context.

Conclusion:
While student choice in data context did not yield higher quantitative performance in this specific study, it successfully transformed the learning experience for a majority of students. It shifted their perception of statistics from a theoretical exercise to a tool for investigating real-world phenomena, thereby increasing intrinsic motivation and the perceived value of the discipline.