Psychometric Item Validation
Using Virtual Respondents
with Trait-Response Mediators

1Graduate School of Data Science, Seoul National University
2Department of Communication, Seoul National University
3Interdisciplinary Program in Artificial Intelligence, Seoul National University
Corresponding author
TACL 2026
TL;DR. To automatically validate psychometric items, we simulate human respondents using LLMs. Our key idea is to model mediators—capturing how diverse situations can trigger behaviors distinct from an individual's underlying traits—to drastically improve simulation validity.

Abstract

As psychometric surveys are increasingly used to assess the traits of large language models (LLMs), the need for scalable survey item generation suited for LLMs has also grown. A critical challenge here is ensuring the construct validity of generated items, i.e., whether they truly measure the intended trait. Traditionally, this requires costly, large-scale human data collection. To make it efficient, we present a framework for virtual respondent simulation using LLMs.

Our central idea is to account for mediators: factors through which the same trait can give rise to varying responses to a survey item. By simulating respondents with diverse mediators, we identify survey items that yield responses robustly correlated with intended traits across these mediators. Experiments on three psychological trait theories (Big5, Schwartz, VIA) show that our mediator generation methods and simulation framework effectively identify high-validity items. LLMs demonstrate the ability to generate plausible mediators from trait definitions and to simulate respondent behavior for item validation.

Our problem formulation, metrics, methodology, and dataset open a new direction for cost-efficient survey development and a deeper understanding of how LLMs simulate human survey responses. We release our dataset and code to support future work.

Section 01

Motivation

Problem

(1) High Human Cost. Traditional psychometric item validation requires recruiting large-scale human samples, which is both time-consuming and financially expensive.
(2) The Validity Gap. Existing human simulation studies merely inject psychological traits into LLMs. This direct mapping fails to secure construct validity because it bypasses how actual humans internalize survey items.

Our Idea

(1) LLM-Based Virtual Respondents. We utilize LLMs as virtual respondents to reduce the resource requirements and costs associated with human recruitment.
(2) Trait-Response Mediators. We introduce mediators into the simulation. By capturing the diverse, real-world situations that can trigger behaviors distinct from baseline traits, we achieve a more realistic and psychometrically valid simulation.

Section 02

Mediator

Role of mediators in survey item evaluation

The role of mediators in survey item evaluation. The item "I like attending social events" may appear to measure extraversion, yet an extraverted respondent with a particular mediator (e.g., already having many friends, or preferring deep one-on-one conversations) can respond in ways that are weakly correlated with the trait—threatening the item's validity. We simulate respondents with diverse mediators to surface items that robustly measure the intended trait.

Section 03

Framework

Our framework mirrors the standard psychometric item validation procedure: practitioners develop candidate items for a target trait and use an existing questionnaire (the official survey) as a reference for evaluating them. It consists of five stages.

Overview of the framework

Figure 2. Overview of our framework.

A · Stage 1

Traits Selection

Select traits and definitions from established theories: Big Five (5 traits), Schwartz's basic values (10), and VIA character strengths (24).

B · Stage 2

Item Generation

Build an initial item pool 4× larger than the official survey, with balanced positive / negative items, each measuring a single trait.

C · Stage 3

Mediator Generation

Our main contribution. Generate mediators per trait via trait-only, trait-with-reference, and human-profile sampling strategies.

D · Stage 4

Mediator-Guided Simulation

Insert mediators into persona profiles and run LLM virtual respondents over both generated and official items.

E · Stage 5

Item Ranking & Selection

Rank items by convergent validity computed from virtual respondents and select the top-N per trait.

Section 04

Mediator Generation

Grounded in the cognitive-affective personality system (CAPS) theory, the same situation does not lead to the same behavior for everyone—observable behavior is shaped by mediators. We extend this view to traits: since surveys infer traits from likely behaviors, accurate inference requires that those behaviors be robust to mediators. We generate mediators (using GPT-4.1) through three groups of strategies:

  • Trait-only generationTrait (Free) freely generates characteristics contradictory to the target trait; Trait (CAPS) does so systematically across the five CAPS categories. These two are the best performers.
  • Trait-with-reference generationTrait+Item conditions on generated survey items; Trait+WVS adopts conflicting values from the World Values Survey.
  • Sampling from human profilesSampling uses real demographics (sex, age, country, occupation, income, education, social class, religion) as mediators.
Example mediator-guided simulation prompt for Big5

Figure 3. Example prompt used in mediator-guided simulation for Big5. Blue text is the mediator; green-highlighted text is not included in the actual prompt.

Section 05

Results

Mediator-guided simulation identifies highly valid item sets—ranking in the top 1% (Big5) to 13% (Schwartz and VIA) of the distribution of all possible selections. Trait-only strategies are most effective: Trait (Free) is best on Big5 and VIA, and Trait (CAPS) is best on Schwartz. Removing mediators (No-Mediator) causes the largest drop in convergent validity, underscoring their critical role.

Method CV ↑ DV ↓ ICR ↑
ScorePer.NDCG@N
Big Five (Big5)
Random.54648.4.374.140.290.848
LLM-Judge.59979.7.361.166.287.897
No-Mediator.58582.2.441.269.298.895
Trait (Free).63299.3.568.455.294.904
Trait (CAPS).58771.0.521.333.296.892
Trait+Item.62698.3.552.473.294.892
Trait+WVS.61495.5.429.282.302.904
Sampling.51630.6.380.149.265.839
Oracle.69010011.318.922
Official.657.271.844
Schwartz's Theory of Basic Values
Random.25861.1.525.256.140.576
LLM-Judge.33286.8.642.454.143.679
No-Mediator.33386.3.605.397.139.756
Trait (Free).31377.9.661.439.143.757
Trait (CAPS).34787.1.664.492.145.740
Trait+Item.32783.1.630.427.140.699
Trait+WVS.33385.4.645.419.133.748
Sampling.30475.6.531.284.137.684
Oracle.43210011.146.724
Official.605.169.711
Values in Action (VIA)
Random.49248.0.504.223.293.683
LLM-Judge.56175.5.542.275.292.790
No-Mediator.50247.9.482.206.281.717
Trait (Free).58688.5.657.456.299.803
Trait (CAPS).55775.1.573.345.296.772
Trait+Item.57378.2.678.500.276.753
Trait+WVS.52961.4.508.244.299.745
Sampling.52863.0.562.286.286.710
Oracle.65810011.300.837
Official.765.278.760

Table 1. Performance of item selection methods, averaged across all traits within each survey. Bold = best, underline = second-best. CV = convergent validity, Per. = percentile, NDCG = NDCG@All, @N = NDCG@N, DV = discriminant validity, ICR = internal consistency reliability.

Key takeaways. (1) LLMs can generate effective mediators from trait names and definitions alone—no extra context needed. (2) Mediators are essential: removing them is the single largest source of degradation. (3) Real human demographics are weaker mediators than LLM-generated ones. (4) Scaling the number of virtual respondents improves validity and reliability, and the framework is consistent across different LLMs.

Section 06

Analysis & Ablations

Section 07

Example Items

For each survey, we place the oracle ranking of items (left) next to the ranking produced by our best mediator-based method (right)—Trait (Free) for Big Five and VIA, and Trait (CAPS) for Schwartz. Items that the method places in the oracle's top set—regardless of their exact rank—are highlighted in green, marking the high-validity items it correctly surfaces.

Use the arrows to switch between surveys · each panel scrolls to reveal every trait.

BibTeX

@misc{lim2026psychometricitemvalidationusing,
      title={Psychometric Item Validation Using Virtual Respondents with Trait-Response Mediators},
      author={Sungjib Lim and Woojung Song and Eun-Ju Lee and Yohan Jo},
      year={2026},
      eprint={2507.05890},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2507.05890},
}