🤖 AI

Gender-Based Heterogeneity in Youth Privacy-Protective Behavior for Smart Voice Assistants: Evidence from Multigroup PLS-SEM

Using multigroup PLS-SEM analysis of survey data from 469 Canadian youths, this study reveals that gender significantly moderates privacy decision-making pathways in smart voice assistant ecosystems, with males showing a stronger link between perceived risks and protective behaviors while females exhibit a stronger connection between algorithmic trust and self-efficacy.

Original authors: Molly Campbell, Yulia Bobkova, Ajay Kumar Shrestha

Published 2026-03-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Molly Campbell, Yulia Bobkova, Ajay Kumar Shrestha

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Always-On" Roommate

Imagine you have a new roommate who never sleeps, never blinks, and is always listening to your conversations just in case you need help. That's a Smart Voice Assistant (like Alexa or Siri).

This study asks a simple but tricky question: Do boys and girls handle the "privacy" of this roommate differently?

The researchers surveyed 469 Canadian young people (ages 16–24) to see how they think about risks, trust, and their own ability to protect their secrets when using these devices. They wanted to know if the "mental recipe" for staying safe is the same for everyone, or if it changes based on gender.

The Main Ingredients (The "Privacy Salad")

To understand the results, imagine privacy decision-making as a salad. The researchers looked at five main ingredients:

Perceived Risk (PPR): How scared you are that the roommate is stealing your secrets.
Perceived Benefits (PPBf): How much you love the convenience (e.g., "It's so easy to set a timer!").
Trust (ATT): Do you believe the company is honest about what they do with your data?
Self-Efficacy (PSE): Do you feel like you have the skills and tools to lock the door and protect yourself?
Protective Behavior (PPB): The actual actions you take, like turning the device off or changing settings.

The Findings: Two Different Playbooks

The study found that while boys and girls generally feel the same amount of fear and have similar levels of trust, they use different mental strategies to decide what to do.

1. The "Direct Action" vs. "Confidence First" Gap

For the Boys (The Direct Route):
Imagine a boy sees a "Danger" sign. His reaction is immediate: "That's risky, I'm going to lock the door right now."
The study found that for males, fear directly leads to action. If they think there is a risk, they are more likely to take protective steps immediately. They don't need to feel "super confident" first; the threat itself is enough to make them act.
For the Girls (The Confidence Bridge):
Imagine a girl sees the same "Danger" sign. She thinks, "That's risky, but do I have the right tools to fix it? Do I trust the system enough to feel safe?"
For females, the path to action is longer. They need Trust first to build Confidence, and then they take action.
- The Analogy: It's like trying to cross a river. Boys might just jump in if they see a crocodile. Girls might first check if the bridge is sturdy (Trust) and if they have a life jacket (Confidence/Self-Efficacy) before crossing. If they don't feel confident, they might freeze, even if they are scared.

2. The "Confidence Gap"

The study found a surprising difference in Self-Efficacy (how confident people feel in their ability to protect themselves).

Boys reported feeling significantly more confident in their ability to manage privacy.
Girls felt less confident.
Why? The researchers suggest this might be because society teaches boys to be "tech wizards" and girls to be cautious. Also, voice assistants often have female voices that sound submissive, which might subconsciously make female users feel less in control.

3. The "Invisible" Groups (Non-Binary & Prefer Not to Say)

The study also looked at a small group of people who didn't identify as strictly male or female. Because the group was small, they couldn't run the full math on them, but the numbers told a sad story:

This group felt the least trust in the technology.
They felt the highest risk.
They felt the least benefit from using the devices.
The Metaphor: If the device is a club, the binary groups (boys/girls) are standing at the door, debating whether to go in. The non-binary group feels like they are standing outside in the rain, feeling unwelcome and unsafe, with no one offering them an umbrella.

Why Does This Matter? (The "So What?")

The researchers argue that one-size-fits-all privacy advice doesn't work.

For Young Men: Since they act on fear, telling them "This is dangerous!" is a very effective way to get them to protect their data.
For Young Women: Since they need confidence first, just saying "It's dangerous" might make them feel helpless. Instead, we need to show them how to protect themselves and build their trust in the system. We need to give them the "life jacket" so they feel capable of crossing the river.
For Everyone: We need to design these devices so that everyone feels like they have control, not just the boys.

The Bottom Line

Privacy isn't just about technology; it's about psychology. This study shows that boys and girls (and gender-diverse youth) navigate the digital world with different maps. To keep young people safe, we need to stop giving everyone the same map and start drawing routes that work for their specific journey.

1. Problem Statement

Smart Voice Assistants (SVAs) are ubiquitous in youth lives, creating unique privacy challenges due to their "always-on" nature and continuous data capture. While existing literature acknowledges gender differences in general digital privacy attitudes (e.g., women often report higher concern), there is a significant gap in understanding how gender shapes the structural pathways leading to privacy-protective behavior specifically within SVA ecosystems.

Previous studies have focused on adult populations or general privacy concerns but have not tested whether the causal relationships between key constructs—Perceived Privacy Risks (PPR), Perceived Privacy Benefits (PPBf), Algorithmic Transparency and Trust (ATT), Privacy Self-Efficacy (PSE), and Privacy Protective Behavior (PPB)—differ between male and female youth. Furthermore, most research excludes non-binary and gender-diverse identities, limiting the inclusivity of privacy models.

2. Methodology

The study employs a quantitative approach using Partial Least Squares Structural Equation Modeling (PLS-SEM) with Multigroup Analysis (MGA).

Data Source: Survey data from 469 Canadian youths (aged 16–24) who use SVAs.
- Groups: Male ( $N=241$ ), Female ( $N=174$ ), Non-binary ( $N=15$ ), and Prefer Not to Say ( $N=39$ ).
- Note: The MGA was conducted only on the binary groups (Male/Female) due to sample size constraints for the other groups.
Model Constructs:
1. Perceived Privacy Risks (PPR): Vulnerability felt when using SVAs.
2. Perceived Privacy Benefits (PPBf): Convenience/advantages offsetting risks.
3. Algorithmic Transparency and Trust (ATT): Belief in developer transparency.
4. Privacy Self-Efficacy (PSE): Confidence in managing/protecting personal info.
5. Privacy Protective Behavior (PPB): Concrete actions taken to safeguard data.
Analytical Procedure:
1. Measurement Model Validation: Confirmed reliability and validity (loadings, AVE, Composite Reliability, HTMT) for both groups.
2. Measurement Invariance (MICOM): Tested for configural, compositional, and scalar invariance. Results showed partial invariance; specifically, PSE and PPBf did not achieve full invariance (unequal means/variances), but compositional invariance was established, allowing for path coefficient comparison.
3. Multigroup Analysis (MGA): Used permutation tests (1,000 subsamples) to compare standardized path coefficients ( $\beta$ ) between males and females.
4. Descriptive Analysis: Reported mean scores for non-binary and "prefer not to say" groups to identify trends without inferential testing.

3. Key Contributions

Structural Pathway Differentiation: This is one of the first studies to apply MGA to SVA privacy, demonstrating that the mechanisms driving protective behavior differ by gender, not just the levels of concern.
Gender Efficacy Gap: Identifies a statistically significant gap where male youth report higher Privacy Self-Efficacy (PSE) than females, even when overall risk perception is similar.
Inclusivity in Privacy Research: Provides preliminary descriptive data on non-binary and gender-diverse youth, highlighting that current binary-focused models may fail to capture the privacy psychology of these groups.
Refined Theoretical Model: Moves beyond the "privacy paradox" (concern vs. action) by showing that the pathway from concern to action is moderated by gender (direct vs. indirect routes).

4. Key Results

A. Construct Mean Differences (RQ1)

Privacy Self-Efficacy (PSE): Males reported significantly higher levels ( $M=3.08$ ) than females ( $M=2.84$ ; $p < 0.01$ ).
Perceived Benefits (PPBf): Males reported marginally higher benefits ( $p < 0.1$ ).
No Significant Differences: Found in Perceived Privacy Risks (PPR), Algorithmic Trust (ATT), or actual Protective Behavior (PPB) between genders. Both groups expressed moderate risk and low trust.

B. Structural Path Differences (RQ2)

The MGA revealed three marginally significant ( $p < 0.1$ ) differences in structural pathways:

Direct Effect of Risk on Behavior (PPR $\to$ PPB):
- Stronger for Males ( $\beta = 0.424$ ) than Females ( $\beta = 0.233$ ).
- Interpretation: Male youth are more likely to translate perceived risk directly into protective action.
Indirect Effect of Trust via Self-Efficacy (ATT $\to$ PSE $\to$ PPB):
- Stronger for Females ( $\beta = 0.229$ ) than Males ( $\beta = 0.132$ ).
- Interpretation: For females, trust in the system is a critical precursor to feeling self-efficacious, which then drives protective behavior.
Mediated Risk Path (PPR $\to$ PSE $\to$ PPB):
- Females: Positive mediation ( $\beta = 0.050$ ). Higher risk perception boosts self-efficacy, leading to action.
- Males: Negative mediation ( $\beta = -0.016$ ). Higher risk perception slightly undermines self-efficacy or operates through a different route.

C. Descriptive Patterns for Gender-Diverse Groups (RQ3)

Non-binary Group ( $N=15$ ): Reported the lowest levels of Trust (ATT) and Benefits (PPBf) and the highest levels of Perceived Risk (PPR) and Protective Behavior (PPB).
Prefer Not to Say ( $N=39$ ): Scores were generally intermediate, aligning more closely with the binary groups.
Implication: Gender-diverse youth may experience a distinct "privacy calculus" characterized by high risk and low trust, necessitating different interventions.

5. Significance and Implications

Design Interventions: Privacy controls should not be "one-size-fits-all."
- For Females: Interventions should focus on building trust and transparency to enhance self-efficacy, as this is the primary driver for their protective behavior.
- For Males: Interventions can leverage risk communication, as they are more likely to act directly on perceived threats.
Policy and Governance: Current SVA interfaces may alienate gender-diverse youth (who show the highest risk/lowest trust). Future designs must move beyond binary gender assumptions to support inclusive privacy management.
Theoretical Advancement: The study confirms that gender acts as a moderator in privacy decision-making models, validating the need for segmented analysis in Human-Computer Interaction (HCI) research.

Limitations: The study is cross-sectional, relies on self-reported data, and is limited to a Canadian sample. Crucially, the non-binary group was too small for structural modeling, highlighting a need for larger, inclusive datasets in future research.