Chapter 4: Self-control

Self-control

Self-control is the problem that we have all desires from ourselves for the long-term that we want to do, but in the short-term, we do very different things.

Why was it so hard for Dan to stick to the treatment regimen and how did he stick to the treatment regiment?

The benefit of the treatment was in the distant future but the costs were immediate.

When he went through the treatment, he created an immediate benefit for his action.

Ego depletion theory (Self-control as a muscle)

Registered replication report (RRR) Only 2 out of 23 labs replicated the effect (self-control model that self-control is like making muscle)

Reward-substitution

Substituting an immediate, tangible reward for a long-term or abstract benefit that may not feel as motivating, making achieving long-term goals more appealing.

What Dan Ariely did for Hepatitis C treatment. Every treatment he went to, he received small rewards (chocolate), which increased his adherence to the treatment regimen.

Self-control contract

Strategy where a person creates a formal agreement—either with themselves or with someone else—to help them stay accountable and follow through on a goal.

Ex. You set up “clocky” — an alarm that would donate money to the charity you hate every 30 seconds you wake up late, to become a 6am morning person.

Ex.2 The Denver Drug program

The Denver Drug Program

An org that makes you write a letter confessing about your drug addiction to the person you don’t want them to know the most (like parents), and they will check your blood time to time to see if you’re still doing drugs. If you get caught, they will send the letter to them immediately.

Study of rats and pigeons pressing buttons to get food: Delayed gratification & commitment

Press the green button → gets food immediately

Press the purple button → gets much more food 10 secs delay

After pressing the purple button, the red button appears. The red button (“no temptation button”) doesn’t give food, but it removes the green button.

Result: Pigeons and rats DO OFTEN press the RED button

Replication crisis

Key factors that got us here

• Publication bias

• Small samples: Watch out for studies with less than 20 ppl per condition (serious skepticism), and proceed with caution if there’s less than 50 ppl per condition

P-hacking: Analyzing the same dataset in a bunch of ways and only disclosing the significant results

• Lazy p-hacking: p’s close to .05 (e.g. p=.04)

• Random covariates: Changing analysis parameters to produce desired results that may not be truly significant.

False positives (incorrectly rejecting the null hypothesis)

A major threat to scientific credibility.

Once published, false positives are hard to eliminate because:

• Replications are rare and often inconclusive.

• Journals rarely publish null results or replications.

Flexibility in data collection, analysis, and reporting inflates the false-positive rate well above 5%, even if researchers think they're following norms.

Simmons et al. (2011) False positive — 4 practices to p-hack and get false-positive

• 2 dependent var. (9.5% FP rate)

• Optional stopping (Collecting more samples until get sig. result)(7.7%) Every few new par. can raise the FP rate up to 22% or more! Very harmful

• Try controlling for gender (11.7%)

• Dropping conditions (12.6%)

Combining all 4 practices → over 60% false positive rate!

Simmons et al. (2011) False positive — Degrees of Freedom

Researcher degrees of freedom = flexibility in:

• Deciding when to stop collecting data.

• Excluding observations.

• Choosing control variables.

• Selecting dependent variables or ways to combine them.

• Researchers often unknowingly justify decisions that lead to significant results due to motivated reasoning.

Key examples of flexibility:

• Varying how "outliers" are treated.

• Picking among multiple possible dependent measures.

• Deciding after the fact how to treat non-significant findings.

Simmons et al. (2011) False positive — study examples that did false positive

Study 1: Feel Older After Listening to Kids’ Music than a control song

Study 2: Feel Younger After Listening to “When I'm Sixty-Four” than a control song

These findings are absurd — but the flexibility in the study design (choosing covariates, flexible data stopping, picking favorable comparisons) made false positives easy to achieve

Simmons et al. (2011) False positive — 6 requirements for authors

1. Pre-specify data collection stopping rules and report them.

2. Collect at least 20 observations per condition unless justified otherwise.

3. List all variables collected in the study.

4. Report all conditions, not just the successful ones.

5. Report results including and excluding any eliminated observations.

6. Report results with and without covariates.

Simmons et al. (2011) False positive — Four Guidelines for Reviewers

1. Enforce the disclosure requirements.

2. Be tolerant of imperfect results.

3. Demand robustness checks (show results under different analytic decisions).

4. Require exact replications if justifications are weak.

Simmons et al. (2011) False positive — Solutions that don’t work

Correcting alpha levels: Impractical because it’s unclear how much to adjust.

Switching to Bayesian statistics: Introduces new degrees of freedom (e.g., choosing priors).

Relying on conceptual replications: Allows researchers to make different choices each time.

Just posting data and materials: Not enough — readers shouldn't have to reanalyze everything themselves.

Duckworth et al. (2018) Increase self-control — theories explaining self-control conflicts

Multiple sequential selves (dynamic inconsistency): Future selves want patience; current selves want immediate rewards

Multiple coexisting selves (competing valuation systems):

• "Planner" vs. "Doer”

• Brain systems for immediate vs. delayed rewards

Multiple attribute models:

• No special "self-control system."

• Choices just involve multiple attributes (like health vs. pleasure).

Duckworth et al. (2018) Increase self-control — 2×2 Framework for interventions

Interventions are classified based on two dimensions:

Situational vs. Cognitive: Targeting the external environment (situational) or internal thought processes (cognitive).

Self-Deployed vs. Other-Deployed: Initiated by the individual (self-deployed) or an external agent like policymakers (other-deployed).

Duckworth et al. (2018) Increase self-control — 4 Self-Deployed Situational Interventions

1. Commitment devices: Pre-commit to future self-control by using tools like “clocky”

2. Temptation bundling: Pair temptations with productive behavior. E.g., Watch TV only when exercising.

3. Situation modification: Remove temptations physically from environment (phone in the backpack instead of on the table)

4. Behavior therapy: Classic behaviorism — Manage environmental triggers and reinforcers

Duckworth et al. (2018) Increase self-control — 7 Self-Deployed Cognitive Interventions

1. Goal setting: Specific, challenging goals outperform "do your best"

2. Planning (Implementation Intentions): "If-then" plans automate behaviors (If I get tired after work, then I hit gym before work)

3. Mental Contrasting and Implementation Intentions (MCII): Combine positive future goal (i wanna get fit) + obstacle confrontation (I get tired after work) + planning (then I go gym before work

4. Self-monitoring: Track behavior to maintain focus

5. Psychological distancing: Adopt third-person perspective to regulate impulses.

6. Mindfulness: Increase nonjudgmental awareness to decouple craving and behavior.

7. Cognitive therapy: Reframe maladaptive thoughts

Alison conducts an experiment on self-control that has 3 conditions, as part of her Honours project. She analyzes her data and finds a small difference between conditions with a p of .08. She is disappointed that she didn’t find a significant difference, so her friends (who are also psych majors) make some suggestions to try to help. Which of these suggestions could increase the likelihood that she will obtain a false-positive result?

a) Try controlling for gender.

b) Try dropping one of the three conditions.

c) Run 10 more people and then conduct the statistical test again.

d) Both a and b

e) All of the above

e) All of the above

One of the most important changes that has happened in the field to combat p-hacking is pre-registration. Pre-registration might be thought of as a way for researchers to overcome a potential self-control failure (i.e., the temptation to p-hack). Duckworth et al (2018) would probably classify pre-registration as a form of…

a) Self-monitoring

b) Goal setting

c) MCII

d) Commitment device

e) Mindfulness

d) Commitment device

In his TEDx talk, Dan Ariely argues that self-control can be enhanced through reward substitution, which he applied to completing his treatment regimen for Hepatitis C. Which of the following studies is designed to test the same idea, that reward substitution promotes self-control?

a) Chocolate lovers come into the lab individually, and all of them are asked to complete extremely difficult math problems. In the experimental condition, the researchers bake chocolate chip cookies, and put the warm, fragrant cookies in front of the participant. Participants are told the cookies are for another study, and they shouldn’t touch them. In the control condition, no cookies are present. The researchers measure self-control by observing how many math problems participants solve before giving up.

b) Chocolate lovers who are planning to take the LSAT (for admission to law school) are all invited to weekly study sessions. Half of them are provided with chocolate at the study sessions and asked not to each chocolate during the rest of the week. Meanwhile, the others are given celery at the study sessions. The researchers measure how many study sessions each participant attends over the course of the semester.

c) University students enrolled in 8am classes are given alarm clocks. They are randomly assigned to receive either a standard bedside alarm clock or a special alarm clock called “clocky,” which runs away from them. The researchers measure self-control by observing how often participants make it to their 8am class on time.

d) Researchers recruit students who want to apply for grad school next fall and are planning to take the GRE (grad school admissions exam) next summer, after spending this school year preparing for it). Half the students are told that the researchers will enter them into a lottery to win a $500 reward if they score anywhere above the 65th percentile on the GRE. Half the students are not told about the lottery. Students’ official scores on the GRE serve as the dependent measure of self-control.

b) is correct. Because providing chocolate as a reward during study sessions encourages self-control by reward substitution for delayed gratification.

a) and d) are wrong because it’s not related to anything we’ve learned at all

c) is clocky, a self-control contract (similar to commitment device)

Simmons et al (2011) note that “The example shown in Figure 2 contradicts the often-espoused yet erroneous intuition that if an effect is significant with a small sample size than it would necessarily be significant with a larger one.” This intuition that we can trust findings based on small samples (as much as larger ones) is most related to what concept you’ve learned so far in 308?

a) Availability heuristic

b) Representativeness heuristic

c) Anchoring-and-adjustment heuristic

d) Group polarization

e) Egocentrism

b) representativeness heuristic

True story: A social psychologist named Diederik Stapel made the news for a study showing that meat-eaters are more selfish than vegetarians. However, it was later discovered that Dr. Stapel made up the data, and even invented the existence of a student who had supposedly worked on the study! It’s now known that Stapel committed similar fraud for dozens of his scientific papers, which have been retracted. Which of the following recommendations are designed to stop people from engaging in behaviour similar to what Stapel did?

a) Authors must decide the rule for terminating data collection before data collection begins and report this rule in the article.

b) Authors must list all variable collected in a study.

c) Authors must report all experimental conditions, including failed manipulations.

d) All of the above.

e) None of the above.

e) none of the above is the correct answer because yes, these three (a, b, c) are designed to prevent questionable research practices (QRPs), and none of these can actually STOP them from committing false-positive fraud.