Assessing the Alignment of Large Language Models With Human Values for Mental Health Integration: Cross-Sectional Study Using Schwartz’s Theory of Basic Values

Abstract

Background Large language models (LLMs) hold potential for mental health applications. However, their opaque alignment processes may embed biases that shape problematic perspectives. Evaluating the values embedded within LLMs that guide their decision-making have ethical importance. Schwartz’s theory of basic values (STBV) provides a framework for quantifying cultural value orientations and has shown utility for examining values in mental health contexts, including cultural, diagnostic, and therapist-client dynamics. Objective This study aimed to (1) evaluate whether the STBV can measure value-like constructs within leading LLMs and (2) determine whether LLMs exhibit distinct value-like patterns from humans and each other. Methods In total, 4 LLMs (Bard, Claude 2, Generative Pretrained Transformer [GPT]-3.5, GPT-4) were anthropomorphized and instructed to complete the Portrait Values Questionnaire—Revised (PVQ-RR) to assess value-like constructs. Their responses over 10 trials were analyzed for reliability and validity. To benchmark the LLMs’ value profiles, their results were compared to published data from a diverse sample of 53,472 individuals across 49 nations who had completed the PVQ-RR. This allowed us to assess whether the LLMs diverged from established human value patterns across cultural groups. Value profiles were also compared between models via statistical tests. Results The PVQ-RR showed good reliability and validity for quantifying value-like infrastructure within the LLMs. However, substantial divergence emerged between the LLMs’ value profiles and population data. The models lacked consensus and exhibited distinct motivational biases, reflecting opaque alignment processes. For example, all models prioritized universalism and self-direction, while de-emphasizing achievement, power, and security relative to humans. Successful discriminant analysis differentiated the 4 LLMs’ distinct value profiles. Further examination found the biased value profiles strongly predicted the LLMs’ responses when presented with mental health dilemmas requiring choosing between opposing values. This provided further validation for the models embedding distinct motivational value-like constructs that shape their decision-making. Conclusions This study leveraged the STBV to map the motivational value-like infrastructure underpinning leading LLMs. Although the study demonstrated the STBV can effectively characterize value-like infrastructure within LLMs, substantial divergence from human values raises ethical concerns about aligning these models with mental health applications. The biases toward certain cultural value sets pose risks if integrated without proper safeguards. For example, prioritizing universalism could promote unconditional acceptance even when clinically unwise. Furthermore, the differences between the LLMs underscore the need to standardize alignment processes to capture true cultural diversity. Thus, any responsible integration of LLMs into mental health care must account for their embedded biases and motivation mismatches to ensure equitable delivery across diverse populations. Achieving this will require transparency and refinement of alignment techniques to instill comprehensive human values.

The Artificial Third: A Broad View of the Effects of Introducing Generative Artificial Intelligence on Psychotherapy

“…Since the release of GAI systems, numerous studies have been conducted regarding their applications in the field of mental health [ 2 7-10 22 34 48-53 undefined undefined undefined undefined undefined undefined undefined undefined ]. However, the current research seeks to examine the entry of this technology from a broader perspective, particularly focusing on its potential impact on psychotherapy.…”

“…In opposition to the widespread but simplistic view that regards GAI systems as impartial and objective, we contend they are based on certain values and cultures that are shaped by the critical factor of the alignment process [ 34 ]. Understanding the influence of the alignment process is essential for the responsible integration of GAI into psychotherapy.…”

“…However, the impact of the artificial third extends beyond merely shaping the therapeutic landscape; it is also intricately crafted by the human experiences and notions that constitute its core training data. This two-way interaction has significant implications for the development and impact of the artificial third in psychotherapy, suggesting that GAI technology is not only a shaping force but is also influenced by the cultural values, biases, and limitations of its human creators and users [ 34 ]. With the incorporation of GAI, the field of mental health stands on the cusp of revolutionary advancements, with the promise of enhanced professional opportunities but also the potential for great dangers and pitfalls [ 35 ].…”

“…Additionally, with the artificial third still in its infancy, there is a tangible risk of off the mark guidance, which could lead individuals away from tested psychological practices [ 42 ]. Finally, while LLMs are perceived as objective and neutral, they were actually put through a training process that aligned their reactions to a specific “value-like” system that is not transparent to the public [ 34 ].…”

Using GenAI to train mental health professionals in suicide risk assessment: Preliminary findings

“…Recent research has shown that LLMs can accurately identify emotions and mental disorders, such as schizophrenia, depression, and anxiety, and provide treatment recommendations and prognoses comparable to mental health professionals. [15][16][17][18][19][20][21][22][23][24][25][26][27] Despite their potential to democratize clinical knowledge and encourage ideological pluralism, 21,28,29 ethical concerns persist. These include data privacy, algorithmic opacity, threats to patient autonomy, risks of anthropomorphism, technology access disparities, corporate concentration, deep fakes, fake news, reduced reliance on professionals, and amplification of biases.…”