Investigate why people act as they do, and what the evidence can establish. Examine an attendance study, design a fair comparison and revise your recommendation when the records change.
Keep a copy of your work. Responses save only in this browser on this device. They are not submitted or synced. Export before changing devices or clearing browser data. Private browsing may not retain your work.
Your work stays in this browser.
Investigate behaviour through evidence
Distinguish an appealing explanation from a claim that a study can support.
What does behavioural science study?
Behavioural science investigates how people make decisions and act within social and material environments. It draws on psychology, economics, sociology and related disciplines. Its questions concern attention, learning, habits, social influence and the relationship between intentions and actions.
The work includes framing a research question, reading previous studies, defining what will count as evidence, designing comparisons, analysing data and explaining what remains uncertain. An intervention can sound persuasive without working; an observed difference can be real without being caused by the intervention.
How the subject is studied
You move between theories, research methods, statistics and critical writing. You learn to operationalise ideas, consider alternative explanations and distinguish a study’s result from its interpretation. Replication, ethics and transparent reporting matter alongside a compelling hypothesis.
What this Lab samples
You will investigate participation in a learning programme. This is an introduction to research design and causal reasoning, not clinical assessment or a test of your ability to diagnose people.
The Next Session investigation
A fictional learning centre wants more enrolled adult learners to attend their next skills session. Its manager believes a planning reminder improves attendance. You will assess the initial comparison, examine a second explanation, propose a better test and respond to a later audit of a follow-up trial.
Your output: an initial judgement, a comparison of explanations, a study protocol, an evidence audit and a research recommendation. All participants in the fictional records are adults aged 18 or over. You will not recruit people, send messages or collect real personal data.
Allow approximately 3–5 hours across one or more sessions at your own pace. No prior psychology study is required. A calculator is useful. Definitions, case evidence and worked examples are included; optional readings extend the investigation.
Calculations, source-supported claims and decisions under the stated rule have checkable answers. Different well-reasoned protocols and explanations can use different wording. Writing is reviewed against explicit criteria; it is not automatically graded.
Examine an apparent improvement
Calculate what happened before deciding why it happened.
Evidence A · The centre’s first comparison
For one session, 80 adult learners chose which of two reminders to receive. Forty opted for a planning reminder and forty chose the standard reminder. They were not randomly assigned. Both reminders contained the same session information and arrived 24 hours before the same session. Neither offered an incentive. Attendance was counted once per person from verified records; no outcomes are missing in this first comparison.
Standard: “Your skills session is tomorrow at 6 pm in Room 2.” Planning: the same information, followed by “Take a moment to decide when you will leave and how you will get here.”
First comparison: distinct learners, one session
Reminder chosen
Learners in group
Learners who attended
Planning reminder
40
28
Standard reminder
40
20
The manager’s claim: “The planning reminder caused a 20-percentage-point increase. We should use it for every learner.”
Learning note: outcome and explanation
Attendance rate = attended ÷ learners in that group × 100. A difference in percentage points subtracts one rate from another. It is not the same as relative percentage change.
Different worked example: 9 of 20 people attend under one arrangement and 6 of 20 under another. Rates are 45% and 30%; the difference is 15 percentage points. That arithmetic alone does not tell you why the groups differ.
A hypothesis is a testable proposed explanation. The planning prompt might help people convert an intention into an action. Alternatively, people already more likely to attend might choose that prompt. Both are possibilities to investigate, not established facts about these learners.
Calculate and form your own initial view
Enter percentages as numbers such as 70 or 70%. For a difference in percentage points, enter the number without a unit. All requested answers are exact whole numbers; do not round a count or replace a missing outcome with an assumption.
In 3–5 sentences, report the difference, assess the manager’s causal claim, give one alternative explanation and identify one additional piece of evidence that would help. Make your first attempt before checking or opening the next evidence packet.
Compare a competing explanation
A combined average can conceal differences in who entered each group.
Record your initial judgement in Step 2 before opening this packet. Your first view will remain in your export when you revise it.
Evidence B · Attendance history before reminder choice
The records below describe the same 80 learners, grouped by attendance at the previous four sessions. “Regular” means attending at least three of those four; “occasional” means attending fewer than three. These labels concern recorded attendance, not motivation, ability or personality. The classifications predate reminder choice.
First comparison split by prior attendance
Prior attendance
Reminder chosen
Learners
Attended next session
Regular attendees
Planning
30
24
Regular attendees
Standard
10
8
Occasional attendees
Planning
10
4
Occasional attendees
Standard
30
12
Compare reminders within each prior-attendance group. Then consider whether the original combined difference persists when both reminder groups are given the same composition.
Learning note: confounding and fair comparison
A confounder offers another explanation for an apparent relationship. Here, prior attendance is related both to reminder choice and to subsequent attendance. Comparing similar prior-attendance groups is informative, but it cannot rule out every unmeasured difference.
For this exercise, also calculate a common-composition rate using 50% regular and 50% occasional learners for each reminder: half the regular-group rate plus half the occasional-group rate. This is a descriptive comparison at a chosen mix, not an estimate proving a causal effect.
Read the correlation-and-causation discussion in OpenStax’s research-methods chapter if you want a broader explanation. The case can be completed from these notes.
Recalculate and revise
Enter percentages as numbers such as 70 or 70%. For a difference in percentage points, enter the number without a unit. All requested answers are exact whole numbers; do not round a count or replace a missing outcome with an assumption.
Explain how group composition generates the initial difference, using two relevant figures. Describe what a reminder-effect explanation and a selection explanation would each predict in a fairer test. State one limitation of using prior attendance alone to make groups comparable.
Design a test that separates the explanations
Write the method before seeing the follow-up result.
The design brief
A second, independent cohort contains 80 consenting adult volunteers from the same centre: 40 regular and 40 occasional attendees, classified using the same earlier-attendance rule. Capacity is sufficient for all 80. All have access to the same reminder channel. Test whether adding the planning sentence changes verified attendance at their next session.
You can allocate individuals to messages, use existing attendance records with coded IDs, and send one reminder 24 hours beforehand. Keep the session, logistical information and support identical. Do not add incentives, deception, public rankings or personal diagnoses. No real study is carried out through this workbook.
Advance decision rule supplied by the centre: recommend a larger evaluation only if the planned comparison was delivered, each person is counted once, every attendance outcome is resolved, and planning-group attendance exceeds standard by at least 10 percentage points. Otherwise investigate the evidence problem or repeat the evaluation. This is a fictional operational threshold, not a statistical-significance test or permission for permanent rollout.
Your proposed protocol
Specify the hypothesis and rival explanation; how allocation works with 40 regular and 40 occasional volunteers; what differs and what stays constant; the primary outcome, denominator and time window; how you will handle duplicate and missing records; and how the stated decision rule applies. Include voluntary participation and coded records.
Write the standard message and your proposed planning version. Both must give the same next-session time and location; add only a neutral prompt to decide when to leave and how to get there. State the hypothesised mechanism and one reason it may not work. Do not promise an effect.
Now use the design audit below to check your method. Your first audit records a snapshot of your protocol before feedback.
Audit the study design
A saved advance plan helps distinguish planned analysis from later exploration. The Centre for Open Science’s preregistration guide explains this distinction. This workbook records a local learning exercise; it does not submit a formal preregistration.
Audit the follow-up before recommending action
A random allocation cannot rescue an outcome count that measures the wrong thing.
Save your protocol in Step 4 first. The later result is a separate supplied trial, not a simulation of your written proposal.
Evidence C · A separate, randomised follow-up
The research team used the reference method: random allocation separately within the 40 regular and 40 occasional volunteers, giving each reminder group 20 regular and 20 occasional learners. Both groups received the same information and timing; only the planning sentence differed. No incentives were used. You are supplied its fictional results, regardless of the exact wording of your own protocol.
The manager’s initial report: “There are 30 attendance scan rows for 40 assigned planning participants and 24 rows for 40 standard participants: 75% versus 60%, a 15-point improvement. We have cleared the 10-point threshold.”
The advance rule from Step 4 still applies. Do not change it after seeing these numbers.
Evidence D · The record audit
Reconciliation of the same follow-up participants
Record
Planning reminder
Standard reminder
Originally assigned learners
40
40
Raw attendance scan rows
30
24
Extra duplicate scan rows
6
0
Distinct confirmed attendees
24
24
Confirmed absentees
12
16
Unresolved attendance outcomes
4
0
The six extra planning rows are second scans by six people already counted as attendees. They are not six additional people. The four unresolved people are separate from the 24 confirmed attendees and the 12 confirmed absentees. Their records were lost; nobody knows from this packet whether they attended. All other records have been reconciled. The three outcome categories partition each assigned group.
Missing is not absent. Keep 40 as the planned denominator. First find the lowest possible planning attendance rate if none of the four unknown people attended; then the highest if all four attended. These are bounds from missing outcomes, not a confidence interval and not bounds on the population causal effect.
Worked example with different numbers
Ten assigned people include 6 confirmed attendees, 2 confirmed absentees and 2 unknowns. Attendance could be 6/10 or as high as 8/10: 60%–80%. Reporting 6/8 answers a different question about resolved records; it does not resolve what the two missing people did.
Reassess the claim
Enter percentages as numbers such as 70 or 70%. For a difference in percentage points, enter the number without a unit. All requested answers are exact whole numbers; do not round a count or replace a missing outcome with an assumption.
Moving this control changes an assumption, not an observed fact. The four outcomes remain unresolved.
Explain which counting error was corrected and what uncertainty remains. Show why the initial 75% claim and a complete-cases-only calculation answer the wrong question. Apply the advance rule and say what could resolve the uncertainty.
Write the research recommendation
Make the progression of your reasoning visible, including what the evidence cannot settle.
Your brief to the centre
Write four short paragraphs: (1) the first comparison and why it did not establish causation; (2) the prior-attendance explanation and the remaining limitation; (3) what the fairer design improved and what the audited follow-up can currently support; (4) your decision under the stated rule and the next evidence needed. Include the relevant rates or differences. Distinguish this centre’s volunteers from all learners elsewhere.
Review the quality of your reasoning
Use group denominators consistently and distinguish percentage points from percentages.
Support the rival explanation with the subgroup pattern; avoid labelling learners as inherently motivated or unmotivated.
Connect each design choice to the inference it improves. Random allocation does not guarantee that a small sample represents everyone.
Preserve the original plan and label later amendments honestly.
Count people once, keep unknown outcomes unknown, and distinguish bounds from a confidence interval.
Recommend a justified next step without treating a pilot score as proof of an effect or a verdict on a career.
Fixed answers receive feedback. Written reasoning uses these criteria. Individual mentor feedback and certificates are not included in this free Lab.
Reflect on the discipline
Choose a specific moment in interpreting data, explaining behaviour, designing a comparison or communicating uncertainty.
Identify evidence that altered or strengthened your view. An accurate initial caution is also a valid response.
Consider a different behaviour, setting, measurement method or competing explanation, and explain why.