Using AI Personas To Explore The Psychological Landscape Underlying The Langer Mindfulness Scale

Intertwining AI And Psychology

As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the advent of modern-era AI that produces mental health advice and performs AI-driven therapy. This rising use of AI has principally been spurred by the evolving advances and widespread adoption of generative AI and LLMs such as ChatGPT, GPT-5, Claude, Gemini, Copilot, and others. For a recap and overview of my well over two hundred analyses and postings about AI and mental health, see the link here and the link here. I frequently speak up about these pressing matters, including in an appearance on an episode of CBS’s 60 Minutes; see the link here.

The use of AI personas is an up-and-coming topic within the field of psychology; see my in-depth analysis at the link here. A budding psychologist or psychiatrist can readily practice or enhance their skills by invoking AI personas that represent different types of human personalities and mental health conditions; see my examples at the link here. This provides a no-harm, no-foul setting that can enable refinement of therapeutic skills. The script can be flipped and invoke an AI persona that acts like a therapist, providing opportunities for understanding what it’s like to be in the shoes of a client or patient; see my discussion at the link here.

Another means of leveraging AI personas consists of performing human psychology research that suitably makes use of this intrinsic LLM capability. You can establish numerous AI personas to be “subjects” in an online experiment or use the AI personas to take tests and surveys. That being said, some would argue that AI personas cannot take the place of human subjects. The use of AI personas can still readily be used in a variety of important ways, which I will describe next.

The Combination Approach

From a methodological perspective, the advent of AI personas provides an added option when it comes to performing psychology experiments.

There are three major pathways typically involved in designing psychological experiments that entail the use of subjects:

• (1) Human-only subjects. The exclusive use of human subjects in an experiment, with no involvement via AI personas.
• (2) AI persona subjects only. The exclusive use of AI personas in an experiment, with no involvement by human subjects.
• (3) Hybrid approach. The use of AI personas and the use of human subjects, such as using AI personas to refine hypotheses, identify notable subgroups, etc., doing so on a pre-test, during, or post-test basis.

One issue is whether you can get “credit” for the AI personas portion of an experiment that is taking a hybrid approach. Some would claim that it is merely preparatory or augmented work. Others would contend that prudent use of AI personas should be given its due, particularly if doing so reveals significant findings. The central differences and similarities to the human subjects’ results could be especially noteworthy.

The Langer Measurement Scale

Shifting gears, I’d like to move into discussing the Langer Mindfulness Scale (LMS). Those of you who are within the field of psychology are likely to be familiar with LMS; others might not know of it. A brief overview will get the topic on the table.

In a now-classic research paper entitled “The Development and Validation of the Langer Mindfulness Scale – Enabling a Socio-Cognitive Perspective of Mindfulness in Organizational Contexts” by Michael Pirson, Ellen J. Langer, Todd Bodner, Sigal Zilcha-Mano, Fordham University Schools of Business Research Paper, October 2012, these salient points were made (excerpts):

• “Mindfulness, understood as a cognitive style, is argued to positively affect various outcomes on the individual level, such as creativity, physical well-being and psychological well-being.”
• “Our study describes the development of a Western understanding of mindfulness and a new socio-cognitive, robust multi-factor and multi-item scale measure of mindfulness.”
• “Using 9 samples (including multi-source and multi-wave data) with a total of 4,345 responses, the authors developed a reliable scale with a tri-dimensional factor structure that replicated across 5 separate samples.”
• “Overall, the present study suggests that the Langer Mindfulness Scale (LMS14) is a reliable and valid measure and that the LMS has important implications for both individuals and organizations.”

Mindfulness tends to involve perceiving things anew. A person is mindful when they avoid a knee-jerk or mindless reaction to the world around them. Their thought processes are considered mindful if they are actively detecting their environment and making new distinctions. The LMS provides a set of fourteen questions that can each be answered on a scoring basis of 1 to 7, producing a total score that reflects a person’s sense of mindfulness.

AI Personas And The LMS

Let’s consider four especially notable ways that AI personas can be used when it comes to the LMS:

• (1) Preparatory. Make use of AI personas as a form of preparation or pre-test in anticipation of using the LMS with human subjects.
• (2) Hypothesis testing. Make use of AI personas to explore novel hypotheses about the nature and scope of mindfulness via the use of the LMS.
• (3) Research instrument review. Make use of AI personas as an investigatory tool to explore the structure and properties of the LMS as a psychological research instrument.
• (4) AI behaviors. Make use of AI personas as targeted test subjects to reveal insights into AI-driven computational and mathematical behaviors by applying the LMS.

I will quickly describe those four possible uses. First, a researcher or practitioner who is aiming to use the LMS on human subjects might want to do some practice runs beforehand. This can be done with ease via AI personas. Trying to obtain human subjects tends to be costly, logistically complicated, and time-consuming, and you ought to be ready to best use their involvement as subjects.

Second, if you have novel hypotheses about mindfulness as a theoretical construct, AI personas are a rapid way to try them out. You can iterate and assess the hypotheses, perhaps subsequently making use of human subjects to explore a curated set of propositions. Third, you can use AI personas to investigate the LMS in its capacity as an important research instrument in psychology, seeking to explore the ins and outs. Fourth, the LMS can provide useful insights about AI itself, namely as a lens into AI-driven computational and mathematical behaviors.

Mini-Experiment With LMS And AI Personas

I will walk you through an example of using AI personas in an LMS context. Consider this a mini-experiment of sorts. I will be doing a series of posts to fully explore this mini-experiment, so stay tuned. Let’s get started with the opening steps.

For my mini-experiment, I have come up with three interesting questions that I’d like to address:

• (a) Do AI personas tend to naturally produce a uniform distribution of LMS scores?
• (b) Do the LMS scores remain relatively stable when the same AI personas are retested?
• (c) Do the LMS scores predict behavior on novelty-detection tasks that were never mentioned when the AI personas were initially created?

This is an example of a synthetic psychometric validation study and provides potentially thought-provoking results.

Creating The AI Personas

By using appropriate prompting techniques, you can craft AI personas in just about any of the major LLMs (see my detailed explanation on how to do so at the link here). Pre-existing AI personas that have been devised by the AI makers are also included in most of the popular LLMs; thus, you can invoke those established AI personas too. When creating your own set of AI personas for psychological research, I’d suggest you consider leveraging my stipulated taxonomy at the link here.

Keep in mind that you can stack the deck when it comes to crafting AI personas. For example, LMS employs three major dimensions, consisting of novelty seeking, novelty producing, and engagement. You could create AI personas that are based solely on the novelty-seeking dimension. The odds are that, upon applying the LMS, doing so would highlight a novelty-seeking propensity. If the researcher didn’t realize they had skewed the AI personas in that direction, they might inadvertently declare that AI personas in general tend to emphasize novelty-seeking behaviors. That’s not fair in this instance, since it was instead a reflection of the way that the AI personas were devised.

The gist is that creating the AI personas should be done with abundant caution and awareness. Try to compose them in a balanced way. Of course, if the purpose of a study is to see what occurs when the deck is stacked, you can certainly steer the AI personas in a particular direction. Just know that you are doing so and make sure to be transparent about what you did and how that impacts the results.

Psychologically Diverse Set Of AI Personas

I went ahead and created a synthetic population of 1,000 psychologically diverse AI personas. The aim was to generate a heterogeneous population. If you wanted to tilt them, you could readily mention that they should be mindful. That would poison the pool in the sense that the AI would then go out of its way to infuse mindfulness into their characteristics. Avoid referring to mindfulness if you don’t want that hint or suggestion to permeate the AI persona population that you are creating.

In my prompt, I indicated that each of the AI personas should have these properties:

• Unique identifier: Included this so that each AI persona can be distinctly reviewed and analyzed.
• Background of the AI persona as crafted by the LLM.
• Personality description of the AI persona as crafted by the LLM.
• Cognitive style of the AI persona as chosen by the LLM.
• Characteristic approach to unfamiliar information by the AI persona.
• Characteristic approach to ambiguity by the AI persona.
• Characteristic approach to problem solving by the AI persona.
• Characteristic approach to changes in circumstances by the AI persona.
• Characteristic approach to generating alternatives by the AI persona.
• Characteristic level of engagement with ongoing situations by the AI persona.

Additional facets could be included, but this is enough for this mini-experiment.

Applying The LMS Instrument

I told the LLM to go ahead and create 1,000 AI personas accordingly. Next, I told the LLM to have each AI persona fill out the LMS and provide their 7-point answers to each question. I then exported the raw data of their answers into a data file, which I imported into a spreadsheet. I also piped the data into a statistical tool.

I had anticipated that a bell-shaped curve would represent their distribution of scores. This seemingly should be the case because I intentionally sought to have a heterogeneous population. I expected a semblance of uniformity. Turns out that there was a skewing toward the middle-to-upper portion of the mindfulness scale.

Why so?

My hunch was that the way in which contemporary LLMs are tuned is entering into the situation. It is a hidden factor. AI makers use RLHF (reinforcement learning with human feedback), see my detailed explanation at the link here, when they are initially tuning their LLMs. They hire humans to give feedback to the AI. The AI then mathematically and computationally adjusts based on that human feedback.

By and large, the humans performing RLHF tend to give a thumbs up to the AI whenever it appears to display curiosity, flexibility, etc. I would suggest that the LMS picked up on that tendency that is now baked into the LLM. The synthetic population appears to be more “mindful” on the LMS than a likely randomly selected human population might be.

More To Explore

The AI personas can be modified or, better still, started anew with a prompt that will tend to overcome the inherent bias of the AI per RLHF. I will do so and tell you how it came out. Also, in the next posting of this series, I’ll be examining the statistical results of the 1,000 AI personas, including the mean, standard deviation, skewness, kurtosis, correlations, and Cronbach’s alpha. I will then proceed to the next step of my mini-experiment, namely having the AI personas undertake a novelty detection task.

Will the LMS score for each respective AI persona be a reliable predictor of how that persona behaves on a novelty detection task? We shall see.

A final thought for now. The famous essayist and poet Paul Valéry made this remark: “The purpose of psychology is to give us a completely different idea of the things we know best.” Generative AI provides an economical way to spur us toward new ideas of the things we know and don’t know about human behavior. Psychology researchers and practitioners should be incorporating AI into their mindset and skill set to aid in further advancing the field of psychology.