Leveraging large language models to assist philosophical counseling: prospective techniques, value, and challenges

Abstract

Large language models (LLMs) have emerged as transformative tools with the potential to revolutionize philosophical counseling. By harnessing their advanced natural language processing and reasoning capabilities, LLMs offer innovative solutions to overcome limitations inherent in traditional counseling approaches—such as counselor scarcity, difficulties in identifying mental health issues, subjective outcome assessment, and cultural adaptation challenges. In this study, we explore cutting‐edge technical strategies—including prompt engineering, fine‐tuning, and retrieval‐augmented generation—to integrate LLMs into the counseling process. Our analysis demonstrates that LLM-assisted systems can provide counselor recommendations, streamline session evaluations, broaden service accessibility, and improve cultural adaptation. We also critically examine challenges related to user trust, data privacy, and the inherent inability of current AI systems to genuinely understand or empathize. Overall, this work presents both theoretical insights and practical guidelines for the responsible development and deployment of AI-assisted philosophical counseling practices.

Full text

ARTICLE
Leveraging large language models to assist
philosophical counseling: prospective techniques,
value, and challenges
Bokai Chen1,5, Weiwei Zheng2,5, Liang Zhao2,3 ✉& Xiaojun Ding 4✉
Large language models (LLMs) have emerged as transformative tools with the potential to
revolutionize philosophical counseling. By harnessing their advanced natural language pro-
cessing and reasoning capabilities, LLMs offer innovative solutions to overcome limitations
inherent in traditional counseling approaches—such as counselor scarcity, difficulties in
identifying mental health issues, subjective outcome assessment, and cultural adaptation
challenges. In this study, we explore cutting‐edge technical strategies—including prompt
engineering, fine‐tuning, and retrieval‐augmented generation—to integrate LLMs into the
counseling process. Our analysis demonstrates that LLM-assisted systems can provide
counselor recommendations, streamline session evaluations, broaden service accessibility,
and improve cultural adaptation. We also critically examine challenges related to user trust,
data privacy, and the inherent inability of current AI systems to genuinely understand or
empathize. Overall, this work presents both theoretical insights and practical guidelines for
the responsible development and deployment of AI-assisted philosophical counseling
practices.
https://doi.org/10.1057/s41599-025-04657-7 OPEN
1Department of Psychology, Academy of Advanced Interdisciplinary Studies, Wuhan University, Wuhan, China. 2School of Information Management,
Wuhan University, Wuhan, China. 3Intelligent Computing Laboratory for Cultural Heritage, Wuhan University, Wuhan, China. 4Department of Philosophy,
School of Humanities and Social Science, Xi’an Jiaotong University, Xi’an, China.
5
These authors contributed equally: Bokai Chen, Weiwei Zheng.
✉email: liangzhao@whu.edu.cn;xiaojunding@xjtu.edu.cn
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 1
1234567890():,;
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Introduction
Philosophical counseling is an emerging discipline that
applies philosophical methods to help individuals navigate
life’s challenges by bridging theoretical concepts with
practical realities (Ding et al., 2024b; Louw, 2013; Savage, 1997).
Unlike traditional psychotherapy, philosophical counseling
focuses on examining unexamined assumptions, values, and
reasoning patterns that may lie at the root of personal dilemmas
(Cohen and Zinaich, 2013). By engaging in dialogue with a
trained philosopher, counselees are encouraged to explore pro-
found questions, gain fresh perspectives on their problems,
challenge existing beliefs, and develop more robust ways of
thinking about their lives and the surrounding world (Gindi and
Pilpel, 2015; Grimes and Uliana, 1998).
While philosophical counseling holds significant potential, it
presently faces several challenges as an emerging field (Amir,
2004; Knapp and Tjeltveit, 2005; Louw, 2013). One notable
challenge is the limited number of trained philosophical coun-
selors, which may restrict access to these services. In addition, the
lack of standardized protocols and the inherently subjective
nature of evaluating counseling outcomes might hinder its growth
and broader acceptance (Knapp and Tjeltveit, 2005). Further-
more, many philosophical counselors may not possess extensive
mental health training, potentially affecting their ability to ade-
quately support counselees with psychological disorders.
Addressing these challenges necessitates the development of
innovative and carefully considered solutions.
In recent years, advanced AI technologies—particularly large
language models (LLMs)—have demonstrated remarkable
potential in natural language processing (NLP) tasks, owing to
their expanded training data, enhanced model architectures, and
exponentially increased parameters (Wu et al., 2024). LLMs have
showcased powerful capabilities in translation, question-answer-
ing, and text generation (Bubeck et al., 2023; Chang et al., 2024)
and have already been successfully applied to complex tasks such
as psychological counseling and education (Fu et al., 2023; Liu
et al., 2024). In the realm of philosophy, LLMs exhibit surprising
ability to generate responses closely resembling those provided by
human philosophers when faced with philosophical inquiries
(Schwitzgebel et al., 2023). Moreover, they possess a degree of
logical reasoning that facilitates the identification of common
logical fallacies (Nutas, 2022). Furthermore, the intuitive, user-
friendly interfaces of LLMs—capable of understanding and
responding in natural language render them valuable tools for
both the general public and researchers (Brown et al., 2020;
Touvron et al., 2023b).
Can LLMs offer entirely new opportunities to enhance philo-
sophical counseling? The answer is promising. Their advanced
language processing and logical reasoning capabilities provide a
strong foundation for integrating them into philosophical coun-
seling. Their complex training process, which leverages enormous
amounts of data including philosophical concepts, enables LLMs
to retrieve necessary knowledge and generate responses during
conversations, thereby creating an impression of “simulated
understanding”for the user. Additionally, the user-friendly con-
versational interfaces of LLMs align well with the dialogical
nature of philosophical counseling. Interest among philosophers
in applying AI tools to philosophy has been noted (Clay and
Ontiveros, 2023). However, despite such enthusiasm, the litera-
ture exploring the intersection of LLMs and philosophical
counseling remains sparse. Notably, only Nutas (2022) discussed
whether GPT-3 meets the fundamental requirements of philo-
sophical counseling, yet this work falls short of further technical
improvements and comprehensive analysis of its capabilities.
Integrating LLMs into philosophical counseling goes beyond
mere compliance with existing requirements. Comprehensive
investigations are necessary not only to establish baseline func-
tionalities but also to shape future expectations. This integration
involves addressing current counseling challenges, enhancing the
efficacy of counseling sessions, and upholding ethical standards.
We must not only investigate whether LLMs can be applied
effectively but also consider the broader implications of their
usage, including both the potential benefits and the problems they
might introduce. Comprehensive research is essential to evaluate
the capabilities of LLMs and their potential to solve real-world
issues, as well as to understand the difficulties that might arise.
Moreover, such research should reflect the latest technological
advancements and incorporate the most relevant techniques
associated with these models. As an initial exploration, this paper
aims to systematically investigate the integration of LLMs into
philosophical counseling by addressing the following three
research questions:
RQ1: How can we technically facilitate LLMs to assist philo-
sophical counseling?
RQ2: What value can LLMs bring to promote better philoso-
phical counseling?
RQ3: What challenges could be encountered when integrating
LLMs to assist philosophical counseling?
The remainder of the paper is organized as follows. Firstly, we
review the development of philosophical counseling and highlight
its current limitations. Subsequently, we introduce three primary
technical approaches for applying LLMs, thereby establishing a
technical foundation for LLM-assisted philosophical counseling.
We then propose the potential value added by LLM assistance in
addressing the current limitations of philosophical counseling.
Finally, we discuss the challenges of integrating LLMs into phi-
losophical counseling—particularly their inability to achieve
genuine understanding and empathy. Through this comprehen-
sive investigation, we argue that while LLMs cannot replace
human counselors, they can serve as powerful tools to extend the
reach and effectiveness of philosophical counseling.
Historical and contemporary perspectives on philosophical
counseling
Evolution of philosophical counseling practices. Philosophical
counseling, as a modern professional practice, has its roots in
both ancient philosophical traditions and in the pioneering work
of contemporary thinkers who sought to reapply philosophical
wisdom to address personal and existential concerns (Amir and
Fatić,2015; Lahav and Tillmanns, 1995). Historically, this prac-
tice can be traced back to Socrates, who engaged Athenians in
dialogues that challenged their assumptions and promoted self-
examination. The Stoics, including Epictetus and Marcus Aur-
elius, as well as philosophers such as Epicurus, further developed
methodologies aimed at achieving a good life through rational
inquiry and ethical living, thereby emphasizing the practical
application of philosophy (Hadot, 1995).
In the modern era, Pierre Grimes emerged as a foundational
figure in philosophical counseling. Beginning in the 1960s in the
United States, Grimes utilized philosophical dialogue—drawing
particularly from Platonic and Neo-Platonic traditions—to help
individuals explore personal dilemmas and achieve self-
understanding. His approach, often referred to as “philosophical
midwifery,”emphasizes guiding individuals to uncover the
underlying beliefs that contribute to their concerns (Grimes and
Uliana, 1998). Grimes’s significant contributions have led to his
recognition as the originator of modern philosophical practice by
the global community of philosophical practitioners, as notably
acknowledged at the International Conferences on Philosophical
Practice (ICPP).
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
2HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Concomitantly, Leonard Nelson and his student Gustav
Heckmann were instrumental in developing the Socratic Dialogue
method in Germany during the early 20th century. Nelson’s work
—focusing on critical philosophy and ethical socialism—pro-
moted collective philosophical inquiry as a means to solve societal
problems and foster democratic thinking (Nelson, 1949).
Heckmann continued this tradition by applying Socratic Dialogue
in educational settings and adult learning environments
(Heckmann, 1981).
Gerd B. Achenbach further advanced the field by establishing
the first formal institution of philosophical practice (Philoso-
phische Praxis) in Germany in 1981 (Achenbach, 1984).
Achenbach advocated a return to the practical roots of
philosophy, distinguishing philosophical counseling from psy-
chotherapy and emphasizing dialogue without predefined meth-
ods or therapeutic goals (Achenbach, 2010). His work
significantly popularized philosophical counseling in Europe
and inspired subsequent practitioners and scholars.
In North America, Lou Marinoff played a key role in bringing
philosophical counseling to public attention with his book Plato,
Not Prozac (Marinoff, 1999), promoting philosophy as a practical
tool for addressing everyday problems. Similarly, Shlomit C.
Schuster contributed to the field through her comprehensive
exploration of philosophical practice as an alternative to
traditional counseling and psychotherapy (Schuster, 1999).
Collectively, these figures have been instrumental in the
resurgence and development of philosophical counseling as a
distinct approach for addressing the complexities of life through
philosophical inquiry and dialogue. By acknowledging their
contributions, we aim to present a comprehensive historical
context that reflects the richness and evolution of philosophical
counseling.
Limitations in current philosophical counseling. Despite its
solid foundation in various philosophical traditions and the sig-
nificant contributions of figures such as Grimes and Marinoff, the
relatively short history of philosophical counseling as a for-
malized practice has resulted in several practical challenges.
While its theoretical depth and potential for addressing complex
life issues are undeniable, traditional philosophical counseling
faces limitations that restrict its accessibility and effectiveness for
a broader audience. These challenges include the difficulty phi-
losophical counselors experience in identifying mental health
issues, the subjectivity in evaluating counseling outcomes, the
scarcity of trained professional philosophical counselors, and
cultural barriers that affect its acceptance and application.
Challenges in identifying mental health issues. Certain issues
addressed within philosophical counseling overlap with mental
health problems. Although there are significant theoretical dis-
tinctions between philosophical and psychological counseling, the
general public often fails to distinguish between the two. This
overlap occasionally leads philosophical counselors to encounter
counselees with genuine psychological disorders. According to
Knapp and Tjeltveit (2005), philosophical counseling can be
broadly categorized into narrow and broad approaches. Regard-
less of the approach, philosophical counselors typically lack the
specialized psychological expertise required to identify mental
health issues that exceed their professional competence.
In narrow-scope philosophical counseling, practitioners focus
on distinctly philosophical problems and usually avoid issues
typically addressed by psychologists. However, counselees’
philosophical concerns do not necessarily guarantee the absence
of mental health issues. For example, individuals experiencing
severe depression or suicidal ideation might seek philosophical
counseling to discuss existential questions about life and death. If
a counselor remains unaware of the underlying psychological
conditions, they may inadvertently overlook signs of depression
that require psychological or medical intervention.
Conversely, practitioners who adopt a broader approach tend
to address a wider array of issues by interpreting philosophical
problems to encompass various mental health concerns that are
not strictly medical or biological. This perspective considers
philosophical counseling as an alternative mental health treat-
ment for individuals dealing with troubled relationships, life
crises, depression, or anxiety (Marinoff, 1999; Schuster, 1999).
However, this assumption is problematic, as most mental illnesses
involve a complex interplay of biological, social, and psycholo-
gical factors (Kinderman, 2005). Philosophical counselors, who
typically lack comprehensive mental health training, may find it
difficult to determine the primary cause of a client’s psychological
issues.
Subjectivity in assessing counseling outcomes. Philosophical coun-
selors have an ethical obligation to empirically demonstrate the
efficacy of their methods (Knapp and Tjeltveit, 2005). However,
evaluations of the effectiveness of philosophical counseling are often
neglected or rely solely on the counselor’s subjective assessment.
Such reliance on subjectivity may stem from a misunderstanding of
scientific methodology and empirical testing (Kreimer and Primero,
2017), leading some to question the legitimacy of philosophical
counseling by labeling it pseudoscientificduetoalackofempirical
evidence (Sivil and Clare, 2018).
To address these concerns, a mixed-methods approach that
incorporates both standardized psychological evaluation tools
and qualitative interviews may provide a robust framework for
assessing outcomes (Tashakkori and Creswell, 2007). Instead of
relying solely on conventional psychological scales—which may
be incongruent with the rationalist foundations of philosophical
counseling—a specialized scale focusing on client satisfaction and
overall well-being could be developed. By emphasizing satisfac-
tion rather than direct efficacy, this approach aligns with the core
principles of philosophical counseling while also addressing the
demand for empirical accountability. Additionally, qualitative
interviews with counselees can capture in-depth, nuanced insights
into their experiences, further enriching quantitative data and
creating a more holistic assessment framework.
While some philosophical counselors may find this combined
approach beneficial, the implementation of mixed-methods
assessments poses its own challenges. Many counselors, primarily
trained in rational and dialogical methods, may lack familiarity
with statistical tools and methodologies, and qualitative inter-
views require significant time and resources to conduct,
transcribe, and analyze.
Scarcity of professional philosophical counselors. Solving life’s
problems and pursuing well-being are universal human needs;
however, the current number of professional philosophical
counselors is clearly insufficient compared to these needs.
Becoming a practitioner in philosophical counseling generally
necessitates a solid theoretical background and profound philo-
sophical knowledge, which in turn requires extensive academic
study and practical experience. Recent trends have indicated a
decline in the number of individuals choosing to study philoso-
phy and obtain relevant degrees (Badola, 2015; National Center
for Education Statistics Database, 2023). In China, for instance,
fewer than one in a thousand university graduates receives a
degree in philosophy (Ministry of Education of the People’s
Republic of China, 2023), thereby limiting the pool of potential
philosophical counselors and exacerbating the scarcity of pro-
fessional talent.
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Although global philosophical cafés have increasingly attracted
participation from both academic and non-academic philoso-
phers (O'Neill and Wang, 2021), philosophical counseling—given
its relatively short formalized history—has not yet achieved the
same level of public recognition and acceptance as psychological
counseling, which enjoys broader familiarity and application.
Consequently, even among philosophy students, awareness of the
methods and existence of philosophical counseling may be
limited, further hindering its dissemination and acceptance. As a
result, individuals seeking philosophical counseling often face
significant challenges in locating a counselor whose expertise
matches their needs.
Cultural barriers and resistance to adoption. To foster the
worldwide development of philosophical counseling, it is essential
to address the cultural differences that shape its reception across
diverse regions. Different countries are characterized by unique
languages and philosophical traditions, meaning that a counseling
approach effective in one cultural setting might not translate
seamlessly to another. Moreover, the use of various native lan-
guages complicates communication and learning between philo-
sophical counselors from different cultures, thereby further
hindering the international exchange of ideas and practices.
For instance, China presents a compelling case study of these
cultural barriers. Its deeply ingrained collectivist culture empha-
sizes emotional restraint; when confronted with psychological
distress or inner conflict, many individuals prefer to suppress
their emotions rather than seek constructive avenues for
expression (Zhang and Yan, 2012). Additionally, the Chinese
cultural value placed on harmonious interpersonal relationships
may discourage individuals from engaging in philosophical
debates, which are sometimes perceived as confrontational (Wei
and Li, 2013). Furthermore, the substantial influence of
Confucianism and Daoism in the Chinese consciousness results
in widely entrenched philosophical frameworks that may not
easily align with conventional Western approaches to counseling
(Hu, 2024).
These cultural factors have constrained the evolution of
philosophical counseling in China. A uniquely Chinese
approach—rooted in the country’s rich cultural and philosophical
heritage—may prove to be more suitable and effective in fostering
acceptance and engagement (Ding and Yu, 2024).
LLMs: pioneering a new era in philosophical counseling
Philosophical counseling represents a distinctive approach that
bridges theoretical perspectives with practical applications to help
individuals overcome life’s challenges. Nonetheless, its wide-scale
implementation is hindered by several critical obstacles. For-
tunately, the advent of LLMs offers a promising avenue for
overcoming these limitations. By leveraging advanced language
processing capabilities, LLMs enable intelligent and user-friendly
services that enhance both the accessibility and effectiveness of
philosophical counseling.
Capabilities and applications of LLMs in philosophy
LLMs and their abilities. The rapid development of artificial
intelligence—particularly through the transformer network
architecture based on attention mechanisms (Vaswani et al.,
2017)—has been transformative. Models such as BERT (Devlin
et al., 2019) and GPT (Radford et al., 2018) have significantly
advanced LLMs over the past five years. Today, LLMs, as
exemplified by GPT, Gemini, Claude, Grok and DeepSeek,
demonstrate extraordinary natural language understanding
(encompassing tasks such as intention recognition and entity
extraction) and natural language generation capabilities, owing to
their vast training datasets and sophisticated technical archi-
tectures (Chang et al., 2024).
Trained on petabyte-scale datasets and utilizing architectures
with billions of parameters, LLMs deliver robust performance in
generating contextually relevant responses to diverse natural
language queries. These abilities make them valuable assets across
various domains, enhancing efficiency and expanding function-
ality. Moreover, techniques like fine-tuning could further improve
their performance on specialized tasks, bolstering natural
language understanding, logical reasoning, mathematical compu-
tation, and alignment with user expectations. For individual users,
ChatGPT’s easy accessibility supports daily activities such as
study, work, and provides both emotional and intellectual
assistance via natural language interaction (Alqahtani et al.,
2023).
Researchers in various disciplines are exploring the integration
of LLMs into domain-specific applications. In particular, the use
of LLMs in psychological counseling and related philosophical
fields paves the way for their adoption in philosophical
counseling. In the realm of philosophy, LLMs show promise as
revitalized pedagogical tools for fostering philosophical dialogue
(Smithson and Zweber, 2024), as they can mimic the discourse
characteristic of human philosophers. Models trained on
philosophical texts can generate responses that are nearly
indistinguishable from those of professional philosophers when
addressing similar questions (Schwitzgebel et al., 2023). Liu et al.
(2024) employed LLMs for Socratic teaching, indirectly demon-
strating their potential to facilitate heuristic dialogues in a
counseling context.
LLMs have also exhibited strong capabilities in mental
health support. They can accurately recognize and respond to
emotional cues, and collaborations between humans and
machines in psychological support have heightened the expres-
sion of empathy—a key component in effective counseling
(Patel and Fan, 2023; Schaaff et al., 2023;Sharmaetal.,2023).
Furthermore, LLMs are capable of performing mental health
assessments (Kjell et al., 2023; Levkovich and Elyoseph, 2023)
and delivering personalized interventions (Blyler and Seligman,
2024a,2024b).
The proficiency of LLMs in philosophical dialogue, mental
health assessment, and intervention underscores their potential
role in advancing these fields. Their integration could render
LLMs invaluable resources in contemporary philosophical
counseling and mental health practices, opening new possibilities
at the intersection of artificial intelligence, philosophy, and
mental health.
Essential requirements for AI-assisted counseling. For LLMs to
function effectively as artificial philosophical counselors, they
must possess several core abilities that align with the foundational
principles of philosophical counseling. These include:
1. Capacity for logical reasoning and philosophical dialo-
gue: LLMs should engage in coherent logical reasoning that
facilitates thoughtful and meaningful philosophical con-
versations with counselees. This requires understanding
complex philosophical concepts and applying them appro-
priately in the context of each client’s concerns (Daniel and
Auriac, 2011; Marinoff, 2002).
2. Recognition and appropriate response to emotional cues:
Although LLMs do not possess true consciousness or
genuine empathy, they should be capable of identifying
emotional cues in a client’s language and responding
sensitively. This form of simulated empathy can help
establish rapport and support the client effectively (Patel
and Fan, 2023; Sharma et al., 2023).
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
4HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

3. Awareness of limitations and referral to human inter-
vention: LLMs should be designed to recognize situations
where a client’s issues exceed their capabilities —such as
severe emotional distress or mental health disorders—and
promptly recommend seeking assistance from a qualified
professional (Knapp and Tjeltveit, 2005; Obradovich et al.,
2024).
4. Facilitation of critical self-reflection: LLMs should assist
counselees in critically examining their assumptions and
beliefs by guiding them through philosophical inquiry. This
process should foster deeper self-understanding and
encourage the exploration of alternative perspectives,
without solely relying on emotional rapport (Cohen and
Zinaich, 2013; Gindi and Pilpel, 2015).
While LLMs can process and generate language that appears to
engage with philosophical concepts or simulate empathy, they
neither truly understand nor authentically empathize. Philoso-
phical understanding—explored across various traditions—trans-
cends mere symbol manipulation based on statistical patterns; it
requires a reflective, interpretive engagement with meaning, a
capacity that non-conscious entities like LLMs lack. Similarly,
their simulated empathy is merely a functional imitation derived
from probabilistic patterns in data.
Despite these limitations, LLMs hold significant potential as
assistants in philosophical counseling. As artificial Socrates, they
can facilitate philosophical inquiry and support human counse-
lors. Although they may not independently perform full-scale
philosophical counseling, LLMs can serve as valuable tools to
augment professional counselors by enriching dialogues, offering
philosophical insights, and streamlining the counseling process.
Technical strategies for integrating LLMs into counseling.
Although LLMs are generally powerful—being pre-trained on
large-scale general knowledge—they may not reach their full
potential in specialized domains without additional training. For
tasks such as philosophical counseling, domain-specific learning
(e.g., fine-tuning on philosophical texts) is essential to further
specialize LLMs (Schwitzgebel et al., 2023).
In this section, we present three mainstream technical solutions
for tuning general LLMs into philosophical counseling assistants,
thereby maximizing their potential. As illustrated in Fig. 1,
traditional philosophical counseling provides both the theoretical
foundation and practical examples upon which LLM-assisted
counseling is built. By following technical pathways such as
prompting, fine-tuning, and retrieval-augmented generation,
LLMs can be endowed with capabilities that are more tailored
to the unique requirements of philosophical counseling, thereby
offering additional advantages.
Prompting. Prompting involves crafting specific prompts to
enhance the performance of LLMs on downstream tasks. This
simple yet efficient method allows for customization of the model
for particular tasks without modifying its underlying structure
(Demszky et al., 2023; PF Liu et al., 2023). By providing a few
training examples as prompts during interaction, the model can
be instructed to perform the desired tasks. The key is to leverage
appropriate task descriptions or examples that guide the model’s
reasoning, enabling it to fully utilize its pre-trained abilities. The
simplicity of prompting rests on designing effective prompt
content—such as annotating the domain and providing illus-
trative examples. Continuous modification and refinement, often
called prompt engineering, are typically required to achieve
optimal results. One widely used approach is the “Chain of
Thought”(CoT) method (Wei et al., 2022); similar to thinking
“step by step,”CoT strengthens the reasoning capabilities of
LLMs by incorporating explicit reasoning steps into the prompts.
Retrieval-augmented generation. Retrieval-augmented generation
(RAG) is an advanced technique that enhances LLM capabilities
by incorporating external knowledge. Although general LLMs are
pre-trained on vast amounts of internet data, they still may lack
sufficient domain-specific knowledge in specialized fields such as
philosophy (Kandpal et al., 2023).
RAG addresses this shortcoming by retrieving relevant
information from a database prior to the generation process,
thereby allowing the model to access a wealth of recent or
domain-specific data. The retrieved information is used to inform
or augment the model’s responses, resulting in outputs that are
more accurate, reliable, and informative (Gao et al., 2024). The
RAG process typically comprises three stages: indexing, retrieval,
and generation. Initially, documents from a knowledge base are
indexed and encoded into vectors. When a query is issued, the
system retrieves the most relevant documents based on semantic
similarity calculations and employs them to generate a compre-
hensive response (Ma et al., 2023). The construction of robust
Fig. 1 Technical framework for integrating LLMs as assistive tools in philosophical counseling (PC).
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 5
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

external knowledge bases is central to RAG, making high-quality,
domain-specific knowledge essential, particularly in the context of
philosophical counseling.
The advantages of RAG include enhanced reliability and
interpretability of domain-specific information, as well as a
greater diversity of retrieved data—which may enrich the
generated responses. However, challenges such as high retrieval
costs and the effective organization of the retrieved information
persist, necessitating further exploration and optimization.
Fine-tuning. Fine-tuning involves adjusting pre-trained LLMs
using relatively small, task-specific datasets to improve their
performance on particular tasks. This process requires preparing
specialized training datasets, sufficient computational resources,
and technical expertise to meet the model’s training needs.
Reinforcement Learning from Human Feedback (RLHF) is fre-
quently employed to further optimize the model’s performance;
in this process, high-quality answers generated by the fine-tuned
model are collected and then used to train a reward model that
guides reinforcement learning.
In the context of philosophical counseling, fine-tuning can be
applied to dialogue models based on existing open-source LLMs,
such as the LLaMA series (Touvron et al., 2023a;2023b). This
approach allows for adjustments in communication style, tone,
and format, thereby enhancing the reliability of the dialogue
output. Fine-tuning is particularly beneficial for tasks that cannot
be effectively addressed through prompting alone, especially
when prompt construction is complex or challenging.
The effectiveness of fine-tuning heavily depends on the quality
of the training corpus. A well-curated corpus enriched with
dialogues on philosophical counseling and mental health care
enables the model to better cater to the emotional and
psychological needs of counselees. It also allows LLMs to adopt
the linguistic and stylistic nuances of philosophical discourse,
resulting in responses that closely align with field-specific
expectations.
Despite its advantages, fine-tuning has notable drawbacks—
chief among them being high training costs. Furthermore,
uploading training data via APIs can escalate expenses, while
local training demands significant computational resources and
technical expertise. Additionally, fine-tuning may lead to issues
such as “catastrophic forgetting”(Kumar et al., 2022), whereby
adjustments to the model’s parameters cause it to lose previously
acquired knowledge.
In this section, we comprehensively compare these three
technical solutions for customizing LLMs as philosophical
counseling assistants, as detailed in Table 1, by evaluating nine
aspects (e.g., implementation, data requirements, effectiveness,
etc.), thus aiding in the practical selection of the most appropriate
approach.
Dominant value of LLM-assisted philosophical counseling
Incorporating LLMs into philosophical counseling is not merely
about automating consultation services; rather, its true value lies
in enhancing the entire counseling process—including pre-
counseling psychological assessments, in-session guidance, and
post-counseling documentation and evaluation. Furthermore,
from a broader perspective, the vast repository of knowledge
contained in LLMs can be adapted to meet the needs of a mul-
ticultural context. These four benefits directly address the major
limitations of philosophical counseling discussed in the Section
“Limitations in current philosophical counseling”, demonstrating
the potential for LLM assistance to improve the effectiveness of
current practices, align philosophical counseling with modern
societal demands, and expand its accessibility to a broader
audience. In this section, we detail the dominant value of LLM
assistance along the following four dimensions.
Mental health assessment and counselor recommendations.
Although LLMs currently cannot replace professional mental
health services (Obradovich et al., 2024), a philosophical coun-
seling recommendation system based on them could be built.
LLMs can be programmed to help identify potential mental
health issues through psychological assessments. Additionally, by
employing techniques such as sentiment analysis, LLMs can
detect signs of mental health challenges from natural language
inputs (Chen et al., 2024; Wankhade et al., 2022). Moreover, their
predictive capabilities enable them to assess mental health states
based on online textual data (Xu et al., 2024).
Beyond identifying mental health challenges, LLMs can assist
ordinary counselees by recommending suitable counselors. The
effectiveness of counseling often depends on the alignment
between a counselor’s approach and a client’s profile including
demographic, clinical, psychological, and cultural factors, as well
as personal preferences (Zhou and Zhang, 2018). A personalized
and adaptable approach is, therefore, essential for optimal
outcomes. In this context, LLMs can facilitate tailored interven-
tions that enhance the overall counseling experience. Fine-tuned
on extensive feedback data, LLMs can help determine the most
appropriate intervention type, whether psychological counseling,
philosophical counseling, or alternative forms of support, based
on the user’s mental health status, preferences, and concerns
(Galitsky, 2024). Crucially, users retain autonomy in selecting
their preferred counseling approach. For instance, if philosophical
counseling is chosen, LLMs can refine recommendations by
suggesting specific modalities, such as Logic-based Therapy
tailored to the individual’s needs (Cohen, 2013; Cohen et al.,
2024). Furthermore, the extensive knowledge base of LLMs and
their capacity for domain-specific training enable them to
comprehend and apply various philosophical frameworks,
positioning them to recommend the philosophical school or
individual philosopher most relevant to a given case.
The proposed framework, depicted in Fig. 2, outlines the
application of LLMs in philosophical counseling systems designed
to effectively direct users to professional mental health services
when necessary. The process initiates with the user’s informed
consent, followed by an initial evaluation of the user’s mental
health status. This evaluation utilizes advanced LLMs to analyze
the counseling issues presented. Based on this analysis, the system
recommends appropriate counseling methodologies tailored to
the user’s specific needs. Subsequently, if the user agrees, the
system suggests compatible counselors who are best suited to
address the identified issues. Additionally, LLMs are employed to
furnish users with carefully selected resources and pertinent
information about available mental health service providers. This
ensures that users with recognized mental health concerns are
directed toward appropriate care.
Crucial to the success of this framework is the algorithm’s
ability to detect potential mental health issues with a high degree
of accuracy, particularly focusing on high recall rates. This
ensures that individuals who may have mental health concerns
are reliably identified and receive the necessary attention, as
emphasized by Rabani et al. (2023). This methodical approach
not only heightens the efficacy of the counseling provided but also
enhances the overall safety and well-being of the user.
Such recognition mechanisms could act as safeguards in long-
term counseling. If a client exhibits significant mental health
problems, the system could facilitate a referral to appropriate
professionals—such as psychologists or psychiatrists—for specia-
lized support, ensuring that philosophical counseling stays within
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
6HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

its scope while collaborating effectively with mental health
services to address counselees’holistic well-being.
Optimized outcome assessment and session summarization.
Traditionally, the assessment of outcomes in philosophical
counseling has either been neglected or overly dependent on the
subjective judgment of counselors, which undermines both
objectivity and consistency. LLMs offer a viable method to
address these issues by facilitating aspects of the evaluation pro-
cess that reduce the impact of human bias. Their advanced nat-
ural language processing capabilities allow them to analyze
conversational data and perform basic statistical analyses (Huang
et al., 2024), thereby providing measurable support and objective
evidence to counter skepticism regarding the effectiveness of
philosophical counseling.
Table 1 Comparative analysis of LLM approaches for enhancing philosophical counseling practices.
Prompting Retrieval-augmented generation Fine-tuning
Key idea Utilizes the LLM’s pre-trained
knowledge without modifying its
structure and parameters.
Enhances philosophical counseling capability by
retrieving information from an external
knowledge base without modifying the model.
Customizes the original LLM using an
in-domain dataset, which involves
updating model parameters.
Implementation •Designing specific prompts to
guide the model in philosophical
counseling.
•Relatively simple prompt design
and optimization.
•Integrating external knowledge sources and
retrieving information prior to generation.
•Requires building an effective knowledge base
and retrieval mechanism.
•Adjusting pre-trained model
parameters using a task-specific
corpus.
•Involves extensive data preparation
and computational resources.
Data required None An external knowledge base containing
philosophical knowledge and semantic
information.
Large philosophy-related datasets.
Customization Low (dependent on prompt design;
limited by pre-trained knowledge)
Moderate (customized through retrieval results
and generation process)
High (requires updating data and
parameters)
Performance
consistency
Varies with prompt quality; may be
inconsistent.
Relatively consistent, depending on the quality
of the knowledge base and retrieval algorithms.
High consistency (dependent on the
quality of the training data and
process).
Cost Low—mainly related to prompt
design and modifications.
Moderate—includes costs for maintaining the
knowledge base and implementing retrieval
mechanisms.
High—involves data collection,
extended training time, and significant
computational resources.
Advantages •Simple and user-friendly.
•No modifications to model
parameters required.
•Provides up-to-date and contextually relevant
information.
•Enhances accuracy and reliability of generated
content.
•Improves task-specific performance.
•Highly adaptable to specific needs.
Disadvantages •Effectiveness depends on prompt
quality.
•Requires continuous adjustment
and optimization.
•Complex implementation and high
maintenance costs.
•Requires efficient coordination between
retrieval and generation processes.
•High training cost.
•Potential issues, such as
“catastrophic forgetting.”
Fig. 2 Architectural blueprint of an LLM-enabled philosophical counseling recommendation system.
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Philosophical counseling is characterized by its methodological
diversity including the “beyond-method”approach, multiple
schools of thought, and a wide-ranging array of client issues
(Ding et al., 2024c; Fatićand Zagorac, 2016; Repetti, 2023). This
diversity makes the uniform application of standardized psycho-
logical scales both impractical and inappropriate. Instead, LLMs
can assist counselors by generating customized questionnaire
items tailored to the specific methodologies and objectives of their
practices, whether as qualitative open-ended questions or
quantitative assessment items. Additionally, these models can
analyze client responses to offer nuanced insights that enhance the
evaluation process and contribute to overall counseling quality.
Beyond simply aiding in outcome assessment, LLMs can also
summarize counseling sessions, as illustrated in Fig. 3. Using
continuous prompt engineering, the LLM is capable of producing
valuable session summaries for both counselors and counselees.
The dialogue and summary shown here are generated using the
GPT-4o model (see Fig. 3). Although LLMs sometimes produce
inaccuracies or “hallucinations”that deviate from the original
content (Adhikary et al., 2024), they can still generate detailed
session summaries. Philosophical counselors can review and
refine these summaries to ensure their accuracy and relevance
before sharing them with counselees. This collaborative process
not only preserves the integrity of session documentation but also
reduces the administrative burden on counselors, thus enabling
them to focus more on the substantive aspects of their practice.
LLMs can additionally offer valuable support in gathering user
feedback post-counseling. Acting as feedback collection assistants,
they can compile preliminary data on client perceptions and
gather open-ended suggestions for improvement. This capability
provides counselors with actionable insights to better inform their
practice. As LLM capabilities continue to evolve, they are
expected to deliver even more sophisticated feedback analysis,
offering innovative perspectives and strategies for refining
counseling methods.
Enhancing accessibility and visibility. The integration of LLMs
into philosophical counseling opens significant opportunities for
enhancing both accessibility and visibility in the field. One
innovative application is the use of LLMs to create digital avatars
for philosophical counselors—a concept that has shown promise
in various contexts (Fink et al., 2024; Oliveira et al., 2024). Fig. 4
outlines the general steps involved in building a digital avatar
based on LLMs. Multimodal technologies could be further
incorporated. With the informed consent of both counselors and
counselees regarding the use of counseling dialogue for AI
training, LLMs can be fine-tuned using individual counselor
dialogue data. This enables the models to simulate a counselor’s
distinctive style, expertise, and approach. Consequently, counse-
lees can explore the methodologies and perspectives of a range of
counselors before making an informed decision on whom to
consult, thereby improving the selection process for professional
support.
These avatars hold significant potential as supplementary tools.
Analogous to the role of LLMs in psychology—where they assist
with mental health support and address common queries (JM Liu
et al., 2023; Na, 2024)—LLM-based philosophical counselor
avatars can serve multiple functions. They can answer philoso-
phical questions, clarify complex concepts, and stimulate
reflective thinking through insightful questioning (Park and
Kulkarni, 2023). For example, these models can be fine-tuned
with specialized conversational datasets targeting specific demo-
graphics, such as children. By generating age-appropriate,
engaging, and thought-provoking dialogues, such interactions
can inspire critical thinking and curiosity from an early age. This
adaptability allows these avatars to address the diverse needs of
different audiences, thereby fostering intellectual growth and deep
philosophical reflection.
In addition to facilitating philosophical dialogue, these avatars
have the potential to generate philosophical counseling responses,
especially given rapid advances in AI technologies. As demon-
strated by Raile (2024) in his exploration of ChatGPT as a
psychotherapist, such applications show considerable promise.
However, this potential must be approached with precaution and
thorough oversight. The responsibility for AI-generated content
ultimately lies with human counselors or supervisory personnel.
Prior to delivering philosophical counseling responses, the AI-
Fig. 3 Example of an LLM-generated session summary derived from counseling dialogues.
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
8HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

generated outputs must undergo meticulous human review to
ensure their accuracy, appropriateness, and safety. Even when the
responses are acceptable, counselors are advised to use them
merely as suggestions or starting points, thus fostering richer and
more meaningful dialogue with counselees. This integration of
human oversight and AI assistance underscores the compelling
potential of LLMs in online philosophical counseling.
A critical advantage of integrating LLMs is their capability to
overcome geographic and financial barriers. As cost-effective,
accessible, and user-friendly platforms, LLMs offer a practical
entry point for individuals who might otherwise lack access to
professional counseling. This is particularly beneficial for those
who are unfamiliar with or hesitant about traditional face-to-face
counseling. For instance, through online platforms, individuals
can interact with LLMs to explore philosophical inquiries,
providing an affordable alternative to in-person consultations.
This technological innovation broadens the reach of philosophi-
cal counseling, making it more inclusive and accessible to
marginalized or economically disadvantaged communities.
Beyond enhancing accessibility, LLMs can help address the
global shortage of professional philosophical counselors. By
embedding expert philosophical knowledge within these models,
LLMs provide scalable and timely support around the clock,
ensuring that individuals in underserved regions can access
meaningful philosophical guidance. In this role, LLMs act as a
bridge—partially compensating for the shortage of trained
professionals while extending the impact of philosophical
counseling.
With user-friendly interfaces and mature applications, inter-
active philosophical counseling services can be widely distributed,
encouraging more people to learn about and experience
philosophical counseling. LLMs could play a crucial role in
raising public awareness of philosophical counseling by lever-
aging digital and social media platforms. Although they cannot
replace the nuanced wisdom of professional counselors, their
ability to captivate and engage users may inspire more individuals
to pursue formal philosophical counseling.
Facilitating cultural adaptation in counseling. LLMs can help
overcome cultural barriers and enhance the adaptability of phi-
losophical counseling across different cultural contexts. They are
particularly advantageous due to their extensive knowledge
repositories (Petroni et al., 2019; Zhu et al., 2023). Even without
specialized training or additional knowledge bases, LLMs can
apply fundamental concepts drawn from various philosophical
traditions to appropriate contexts. This capability is illustrated in
a hypothetical dialogue between a GPT-4-based philosophical
counselor and a counselee, where the LLM accurately incorpo-
rates key ideas—such as filial piety and righteousness from
Confucianism—to assist in resolving the counselee’s dilemmas
(Ding et al., 2024a).
As depicted in Fig. 5, by constructing comprehensive knowl-
edge bases of philosophical ideas from different cultures and
employing RAG technology, LLMs can deliver tailored insights.
For example, a complete database of Confucian and Daoist
philosophies can help counselors better understand the cultural
backgrounds and behavioral influences of their Chinese
counselees.
Additionally, LLMs can be utilized for machine translation and
even simultaneous interpretation, which facilitates effective
communication between counselors and counselees speaking
different native languages (Wang et al., 2024; Zhang et al., 2023).
Furthermore, LLMs possess multilingual capabilities that enable
them to simulate the roles of counselees from diverse cultural
backgrounds. By applying prompt templates enriched with
demographic information, cultural context, linguistic styles, life
experiences, and client identity, LLMs can, to some extent,
emulate counselees from varying cultural backgrounds and life
histories. This feature is particularly valuable in multicultural
societies, helping philosophical counselors to better engage with
clients from diverse cultural and experiential backgrounds.
Through these mechanisms, LLMs can significantly enhance
the capacity of philosophical counselors to navigate cultural
complexities, thereby providing robust support that fosters a
more inclusive and effective counseling environment.
Fig. 4 Process for constructing digital counselor avatars using LLMs-related techniques.
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Fundamental challenges of deploying LLMs in philosophical
counseling
While LLMs have shown considerable potential as valuable
assistants in philosophical counseling by providing multi-faceted
support, their application is not without significant limitations.
Despite their remarkable capabilities, LLMs face inherent chal-
lenges when deployed in philosophical counseling—especially in
meeting the comprehensive requirements of such services. These
challenges stem not only from the complexities of
human–computer interaction but also, more critically, from the
fundamental disparities between the nature of philosophical
counseling and the underlying mechanisms of contemporary AI
technologies. Consequently, these limitations demand careful
examination to ensure that LLMs serve as a complementary
resource that augments, rather than inadvertently compromises,
the integrity of the counseling process.
Trust. The integration of LLMs into philosophical counseling
holds the potential to provide effective support for counselors.
However, it also introduces critical challenges, particularly in
terms of human–AI trust. Trust could be considered from mul-
tiple perspectives: counselor trust in machines, client trust in
machines, and the public’s confidence in AI.
A primary concern is the inherent lack of explainability in
LLMs. Their operational mechanisms are complex and often
insufficiently transparent, making it difficult for counselors to
understand the reasoning behind generated outputs and thus
trust their conclusions (Balasubramaniam et al., 2023; Zhao et al.,
2024). Even when LLMs generate a response, they may not
provide explanations that align with human logical processes
(Turpin et al., 2023). This “black box”nature can lead to
skepticism—particularly in philosophical counseling where issues
often require nuanced interpretation—and might ultimately
undermine the confidence of both counselors and counselees,
thereby compromising the quality and impact of counseling
services.
Another challenge concerns the instability of LLM perfor-
mance, which primarily affects client trust. Research indicates
that the same model can produce inconsistent results across
different runs—with accuracy variations reaching up to 10% (Atil
et al., 2024). Such instability may result in biased or even harmful
outputs. For example, there have been reports of LLMs (e.g.,
Google’s AI chatbot) generating inappropriate and potentially
harmful suggestions (CBS News, 2024). In the context of
philosophical counseling, where precision, ethical sensitivity,
and trust are fundamental, these performance fluctuations
present a significant barrier.
Despite these challenges, the advanced capabilities of LLMs
have fostered public optimism. A recent study revealed that
therapists were often unable to reliably distinguish between
transcripts of human–AI interactions and those of
human–human therapy sessions (Kuhail et al., 2024), and surveys
indicate that 55% of respondents are optimistic about the
potential of AI in mental health contexts (Aktan et al., 2022).
However, such optimism also carries risks. Counselors who
uncritically adopt LLM-generated suggestions may inadvertently
undermine their own professional judgment, and when errors
occur, this dependency can amplify the impact of mistakes—
compromising both the professionalism and credibility of the
counseling process.
Addressing these trust-related issues requires striking a careful
balance between leveraging the benefits of LLMs and implement-
ing robust oversight, ensuring that their usage remains transpar-
ent and ethically sound—while preserving the primacy of human
judgment.
Privacy. Privacy is a fundamental aspect of both traditional
counseling and the emerging use of LLMs. In conventional face-
to-face counseling, professionals—whether psychologists or phi-
losophers—are obligated to safeguard counselees’personal
information. However, the shift to online counseling introduces
new vulnerabilities. Although digital communication enhances
accessibility, it also raises the risk of data interception and leakage
(Kiriakaki et al., 2022). These risks become even more pro-
nounced when employing LLMs in counseling settings, as user
inputs may be used to train these models, thereby raising sig-
nificant ethical and privacy concerns. While organizations such as
OpenAI assert that models like ChatGPT do not collect or dis-
close personal information from interactions, numerous reports
of privacy breaches suggest that these safeguards may not be as
reliable as claimed (Yao et al., 2024).
Moreover, unlike psychological counseling—which focuses
more on emotion—philosophical counseling often involves
dialogues about values, life meaning, and ethical dilemmas. This
Fig. 5 Leveraging RAG for cultural adaptation in counseling contexts.
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
10 HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

type of data, although not as directly sensitive as medical or
mental health data, can include deeply personal information such
as an individual’s religious beliefs, moral stances, or personal
philosophies. The exposure of such data could result in significant
repercussions, including social prejudice and discrimination.
These distinctive privacy challenges underscore the importance
of tailoring privacy safeguards to the specific context of
philosophical counseling. By recognizing and mitigating privacy
risks—through measures such as targeted policy regulation and
data protection protocols—we can enhance the effectiveness of
LLMs while ensuring that users’confidentiality is maintained,
thereby fostering trust and security in digital counseling
environments.
Philosophical understanding and empathy. Philosophical
counseling requires a profound level of understanding that goes
beyond mere language comprehension. Counselors must grasp
counselees’issues within their unique personal and cultural
contexts, maintain a deep understanding of philosophical con-
cepts, and—as a critical component—demonstrate empathy, the
capacity to understand and share another person’s feelings
(Cooper and McLeod, 2010). Empathy, as a cornerstone of
counseling, enables deeper engagement with clients’concerns and
fosters a meaningful exploration of ethical, existential, and per-
sonal dilemmas.
To fully appreciate the role of empathy in philosophical
counseling, it is essential to engage with its philosophical
underpinnings. David Hume (1739/2000) posited that empathy
—or “sympathy,”as he termed it—involves an affective resonance
with others’emotions that is inherently human and extends well
beyond mere cognitive simulation. Hume argued that sympathy is
not merely an intellectual exercise but a deeply emotional
connection grounded in our shared humanity. This emotional
dimension underscores a critical limitation of AI: its lack of the
embodied experience necessary for genuine empathy.
Max Scheler (1913/1970) further emphasizes empathy as a
relational and intentional act. In his phenomenological frame-
work, empathy (Einfühlung) is distinguished from mere emo-
tional contagion; it requires engaging with the Other as a subject,
recognizing their unique perspective and lived experience—a
relational depth that LLMs, operating solely on syntactic and
algorithmic principles, cannot replicate. Similarly, Michael Slote
(2007) underscores empathy’s moral significance, presenting it as
central to ethical understanding and deliberation. Slote’s ethic of
care stresses that moral reasoning is not solely an abstract logical
exercise but is deeply rooted in empathic engagement—an
attribute that remains out of reach for LLMs due to their lack
of genuine emotional capacity.
These philosophical insights collectively highlight the human-
centered nature of empathy and its indispensable role in
philosophical counseling. In this context, empathy is not merely
a supportive skill but a core component of philosophical dialogue,
particularly in addressing moral and ethical dilemmas. Although
recent LLMs such as Grok 3, OpenAI o3, Gemini 2.0, Claude 3.5,
and DeepSeek-R1 demonstrate impressive abilities in processing
and generating coherent, contextually relevant language, their
operation remains confined to syntactic processing rather than
achieving true understanding (Pearl, 1988; Searle, 1980). In
simulating understanding through statistical patterns in data,
LLMs fundamentally lack the embodied human experiences,
emotional resonance, and intentionality—qualities that are
critical for genuine philosophical engagement (Boden, 1998;
Bengio et al., 2003).
John Searle’s Chinese Room Argument (1980) powerfully
critiques the claim that AI can truly understand natural language.
In this thought experiment, an individual inside a room follows
syntactic rules to respond to Chinese characters without
comprehending their meaning. This scenario mirrors how LLMs
function: they manipulate symbols based on rules (syntax)
without grasping meaning (semantics). In the realm of philoso-
phical counseling, this distinction is pivotal. Although LLMs can
generate responses that appear meaningful and contextually
appropriate, their outputs are fundamentally probabilistic pre-
dictions rather than intentional expressions of meaning or
emotion—as would be expected from a human counselor
(Harnad, 1990; Ringle, 2019).
Empathy, in particular, underscores this limitation. It is not
only about recognizing emotional cues but also about forming an
authentic emotional connection with the client. Currently, LLMs
lack consciousness and genuine emotional experience, meaning
they cannot truly empathize with counselees. They can produce
empathetic-like responses derived from patterns in their training
data—useful for recognizing distress or offering comforting
language—but these responses do not match the profound
empathic engagement that human counselors can provide
(Chaturvedi, 2024; Elliott et al., 2011). For example, Rubin
et al. (2024) stress that while AI systems may mimic empathic
expressions, they are unable to replicate the intentionality,
concern, and trust-building necessary for meaningful therapeutic
relationships.
The embodied and situated nature of human understanding
further reinforces the limitations of LLMs. Scholars like Hubert
Dreyfus (1972) and Sherry Turkle (2011) argue that human
intelligence is deeply entwined with our physical embodiment
and lived experience, which critically shapes our ability to
navigate complex interpersonal and ethical domains. AI systems,
lacking this embodied perspective, are therefore inherently
incapable of addressing the existential and contextual dimensions
of human concerns—a critical component of philosophical
counseling. Shteynberg et al. (2024) similarly note that LLMs
are currently incapable of genuine empathy, potentially adversely
affecting users seeking emotional connection in their
communications.
Nevertheless, the limitations of LLMs do not render them
irrelevant in philosophical counseling. Instead, their capabilities
should be viewed as assistive and complementary to human
counselors, not as replacements. LLMs can organize and present
philosophical insights, identify pertinent theories, or generate
preliminary analyses of ethical dilemmas. However, the deeper
tasks of understanding a client’s personal context, engaging in
moral reasoning, and providing meaningful guidance necessitate
the uniquely human capacities of judgment, empathy, and
awareness (Guo et al., 2024; Lee et al., 2021; Xu et al., 2024). In
this sense, while LLMs may serve as effective tools for supporting
philosophical inquiry, they cannot substitute the relational and
empathetic depth that truly defines philosophical counseling.
Discussion
Philosophical counseling serves as a bridge between the general
public and complex philosophical theories, making abstract
concepts more accessible and applicable to everyday life. In this
context, AI emerges as a powerful tool to strengthen this con-
nection. The rapid evolution of AI technologies, particularly
LLMs, has showcased their transformative potential across var-
ious domains—including philosophy. Integrating LLMs into
philosophical counseling represents a particularly promising
advancement with the potential to address practical limitations
faced by traditional counseling methods.
Research indicates that LLMs have the capacity to fulfill several
foundational requirements of philosophical counseling (Nutas,
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 11
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

2022). For instance, they can articulate the ideas of major Eur-
opean philosophical thinkers, detect logical fallacies, and elucidate
the role of applied philosophy. However, significant limitations
remain. In particular, LLMs inherently lack true understanding or
empathy—as highlighted by Searle’s Chinese Room Argument,
which distinguishes between syntactic processing and genuine
semantic comprehension. In a field where deep personal
engagement and empathy are critical, this limitation necessitates
careful consideration in their application.
The inherent shortcomings of current AI systems—especially
regarding understanding and empathy—stem from the funda-
mental principles of their design. Achieving AI systems with logic
and emotions truly aligned with human capacities still requires
substantial technological advancements. Striking a balance
between leveraging LLMs’strengths and acknowledging their
limitations is essential for responsible implementation in philo-
sophical counseling. Consequently, LLMs should be viewed as
complementary tools, aiding human counselors in technical and
procedural tasks rather than replacing the nuanced under-
standing and emotional connection that only human practi-
tioners can provide.
From the viewpoint of their role as assistants, LLMs can
enhance philosophical counseling by supporting both counselors
and counselees before and after sessions. They can provide
recommendations, evaluate outcomes, and facilitate cultural
adaptation. The integration of LLM-related techniques—includ-
ing prompt engineering, RAG, and fine-tuning—can evolve phi-
losophical counseling into a more efficient and accessible service.
When combined with other AI technologies such as multimodal
models and recommendation systems, LLMs may further amplify
counselors’capabilities in addressing diverse client needs. Addi-
tionally, features such as 24/7 availability and easy accessibility,
which are not always practical for human counselors to maintain,
make LLMs particularly valuable in broadening the reach and
impact of philosophical counseling services.
While true philosophical understanding and genuine empathy
remain challenging for AI, practical applications may not require
these profound capacities. If AI systems can exhibit behaviors that
resemble human understanding and empathy—akin to Searle’s
description of language simulation in the Chinese Room—they
may still deliver significant practical value. Studies such as Kuhail
et al. (2024) suggest that advanced LLMs can produce responses
closely resembling those of human interactions, potentially
meeting user expectations. Nevertheless, further evidence and
improvements in model architecture and LLM-related techniques
are needed to validate these claims and convincingly emulate
human philosophical counselors.
In addition to the challenges of understanding and empathy,
LLMs face significant obstacles in philosophical counseling,
including issues of human–AI trust, privacy, and security.
Building trust in AI requires transparency in system design, rig-
orous testing to ensure reliability, and clear communication of
model limitations. Privacy concerns can be mitigated through
robust data encryption, anonymization, and adherence to strict
ethical guidelines for data use. Similarly, advanced security
measures—such as secure architectures and real-time threat
detection—are essential to safeguard sensitive client information.
A comprehensive, interdisciplinary approach is indispensable to
overcome these challenges and develop AI systems that are not
only effective but also trustworthy and secure.
AI represents a transformative technology reshaping nearly
every facet of human life, with its influence on philosophical
counseling becoming increasingly apparent. The integration of
LLMs into this domain presents a compelling and promising
avenue of exploration, despite the complexities associated with
applying computational methodologies to a traditionally human-
centered practice. The development of an AI-driven philosophical
counselor or practitioner could manifest as a specialized digital
agent, distinct from existing general-purpose models such as
ChatGPT or DeepSeek-R1. Ideally, such a model tailored speci-
fically for philosophical practice would exhibit advanced profi-
ciency in both philosophical reasoning and counseling
methodologies, surpassing human practitioners in certain
domains while ensuring critical human oversight in value-laden
decision-making processes.
This paper proposes practical strategies for leveraging LLMs as
assistants in philosophical counseling to make abstract philoso-
phical concepts more accessible and to address the limitations of
traditional counseling practices. By introducing and evaluating
three key LLM-related techniques, the study highlights their
respective advantages and disadvantages. Based on these techni-
ques, detailed implementation plans for using LLM technologies
are presented, offering actionable solutions to specific counseling
challenges. Furthermore, this paper critically examines the lim-
itations and challenges associated with LLM-based philosophical
counseling, providing a balanced assessment of its potential and
future directions. In doing so, it contributes to the inter-
disciplinary discussion on whether philosophical counseling will
evolve into a robust field or be dismissed as pseudoscience
(Kreimer and Primero, 2017).
However, this paper has its limitations. It is primarily theore-
tical and prospective in nature, lacking empirical validation or the
implementation of a fully functioning LLM-based system. While
we explore how LLM-related technologies can enhance philoso-
phical counseling, developing a model that simultaneously
addresses all requirements remains a formidable challenge.
Moreover, we do not provide ready-made solutions for deeper
philosophical and ethical obstacles related to AI’s role in coun-
seling. Future research should prioritize the practical imple-
mentation of LLM technologies and conduct empirical studies to
assess their effectiveness in real-world counseling scenarios.
Additionally, ongoing exploration is required to resolve the
challenges identified in this paper and to refine the integration of
LLMs into philosophical counseling.
We acknowledge that many of the issues raised by the
reviewers reflect genuine challenges in balancing the inter-
disciplinary scope of our work with a firm philosophical stance.
Our focus on predominantly Western scholarly sources has left
non-Western contributions—such as those from the Korean
philosophical counseling community and other informal practices
—relatively underexplored. The embodied and contextual nature
of human cognition remains critical to understanding philoso-
phical practice (Shapiro, 2010; Wilson, 2002). Future research
could aim to integrate diverse cultural perspectives and embodied
approaches—drawing, for example, on the insights of Varela,
Thompson, and Rosch (1991)—to address the multi-faceted
dimensions of philosophical counseling, including its historical,
cultural, and embodied underpinnings, in a more comprehensive
manner.
Moreover, we wish to clarify our stance on the role of LLMs in
philosophical counseling. We fully share the reviewers’concerns
regarding the potential risks of overreliance on computational
methods, which might lead to a homogenization of counseling
practices or marginalize the unique contributions of human
counselors. It is crucial to emphasize that our proposal does not
aim to supplant human practitioners but to serve as a supple-
mentary tool—enhancing accessibility and efficiency while pre-
serving human critical judgment. Concerns raised by Searle’s
(1980) Chinese Room argument and further explored by Guo
et al. (2024) underscore the challenges of replicating genuine
understanding and empathy through AI. Future investigations,
perhaps building on empirical studies like those by Kuhail et al.
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
12 HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

(2024), should rigorously assess the impact of AI-assisted inter-
ventions on independent thought and subjectivity in counseling.
We advocate for frameworks that ensure AI augmentation
remains complementary, with robust mechanisms for human
oversight, iterative feedback, and the preservation of the coun-
selor’s unique expertise.
Finally, we recognize that the integration of advanced com-
putational tools into philosophical counseling carries profound
societal and ethical implications. As we further develop technical
solutions—such as fine-tuning, retrieval-augmented generation,
and refined prompt engineering—future research must also
examine whether these interventions might inadvertently shift
philosophical practice toward a consumptive rather than reflec-
tive mode. It is imperative to foster interdisciplinary collaboration
among philosophers, cognitive scientists, and AI researchers not
only to enhance methodological precision but also to ensure that
the core humanistic and transformative values of philosophical
counseling are respected and preserved (Boden, 1998; Slote,
2007). By rigorously interrogating the balance between techno-
logical innovation and the preservation of critical, embodied, and
culturally inflected thinking, we hope to expand a research agenda
that addresses these complex challenges while remaining true to
the transformative aims of philosophical inquiry.
Conclusion
LLMs exhibit significant promise for enhancing philosophical
counseling by addressing key challenges—ranging from limited
service accessibility to subjective evaluation criteria. Yet, inte-
grating these advanced systems within an intrinsically human and
value-laden domain mandates a cautious, balanced approach.
Recognizing that current AI models lack genuine understanding
and emotional empathy, LLMs should serve as sophisticated
adjuncts, rather than replacements, for human counselors. By
augmenting counseling processes and enabling personalized,
scalable interventions, LLMs can help drive a transformative shift
in philosophical practice. Ultimately, with robust oversight,
stringent privacy safeguards, and continuous technical refine-
ment, the responsible adoption of LLM-assisted methodologies
will foster personal growth, enhance ethical discourse, and pro-
mote inclusive, culturally sensitive counseling in an increasingly
digital world.
Data availability
The datasets generated and analysed during the current study are
available from the corresponding authors upon reasonable
request.
Received: 1 August 2024; Accepted: 24 February 2025;
References
Achenbach GB (1984) Philosophische praxis. Verlag für Philosophie Jürgen Dinter,
Cologne
Achenbach GB (2010) Introduction to philosophical practice: lectures, essays,
conversations, and essays presenting philosophical practice from 1981 to
2009. Verlag für Philosophie Jürgen Dinter, Cologne
Adhikary P, Srivastava A, Kumar S et al. (2024) Exploring the efficacy of large
language models in summarizing mental health counseling sessions: bench-
mark study. JMIR Ment Health 11:e57306, https://mental.jmir.org/2024/1/
e57306
Aktan ME, Turhan Z, Dolu I (2022) Attitudes and perspectives towards the pre-
ferences for artificial intelligence in psychotherapy. Comput Hum Behav
133:107273. https://doi.org/10.1016/j.chb.2022.107273
Alqahtani T, Badreldin HA, Alrashed M et al. (2023) The emergent role of artificial
intelligence, natural learning processing, and large language models in higher
education and research. Res Soc Adm Phar 19:1236–1242. https://doi.org/10.
1016/j.sapharm.2023.05.016
Amir LB (2004) Three questionable assumptions of philosophical counseling. Int J
Philos Pract 2(1):58–81. https://doi.org/10.5840/ijpp2004214
Amir L, FatićA (2015) Practicing philosophy. Cambridge Scholars Publishing,
Newcastle upon Tyne
Atil B, Chittams A, Fu L, Ture F, Xu L, Baldwin B (2024) LLM stability: a detailed
analysis with some surprises. https://arxiv.org/abs/2408.04667
Badola BP (2015) Decline of philosophy in academic discourses: a problem with
the modern trends in knowledge. Quest J UGC-HRDC Nainital 9(3):209–215.
https://doi.org/10.5958/2249-0035.2015.00031.5
Balasubramaniam N, Kauppinen M, Rannisto A, Hiekkanen K, Kujala S (2023)
Transparency and explainability of AI systems: from ethical guidelines to
requirements. Inf Softw Technol 159:107197. https://doi.org/10.1016/j.infsof.
2023.107197
Bengio Y, Ducharme R, Vincent P, Janvin C (2003) A neural probabilistic language
model. J Mach Learn Res 3:1137–1155
Blyler AP, Seligman ME (2024a) AI assistance for coaches and therapists. J Posit
Psychol 19(4):579–591. https://doi.org/10.1080/17439760.2023.2257642
Blyler AP, Seligman ME (2024b) Personal narrative and stream of consciousness:
an AI approach. J Posit Psychol 19(4):592–598. https://doi.org/10.1080/
17439760.2023.2257666
Brown T, Mann B, Rtder N et al. (2020) Language models are few-shot learners. In:
Proceedings of the 34th International Conference on Neural Information
Processing Systems (NIPS). Curran Associates Inc., pp. 1877–1901
Boden MA (1998) Creativity and artificial intelligence. Artif Intell 103(1-
2):347–356. https://doi.org/10.1016/S0004-3702(98)00055-1
Bubeck S, Chandrasekaran V, Eldan R et al. (2023) Sparks of artificial general
intelligence: early experiments with GPT-4. https://arxiv.org/abs/2303.12712
CBS News (2024) Google AI chatbot responds with a threatening message: “Human
…Please die.”Available online: https://www.cbsnews.com/news/google-ai-
chatbot-threatening-message-human-please-die/. Accessed on 28 Dec 2024
Chaturvedi A (2024) Exploring empathy in artificial intelligence: synthesis and
paths for future research. Inf Discov Deliv. https://doi.org/10.1108/IDD-03-
2024-0048
Chang YP, Wang X, Wang JD et al. (2024) A survey on evaluation of large language
models. ACM Trans Intel Syst Tec 15(3):1–45. https://doi.org/10.1145/3641289
Chen Z, Deng JW, Zhou JF, Wu JCZ, Qian TY, Huang ML (2024) Depression
detection in clinical interviews with LLM-empowered structural element
graph. In: Proceedings of the 2024 Conference of the North American
Chapter of the Association for Computational Linguistics: Human Language
Technologies (Volume 1: Long Papers). ACL, pp. 8181–8194
Clay G, Ontiveros C (2023) Philosophers ought to develop, theorize about, and use
philosophically relevant AI. Metaphilosophy 54(4):463–479. https://doi.org/
10.1111/meta.12647
Cohen ED (2013) Theory and practice of logic-based therapy: integrating critical
thinking and philosophy into psychotherapy. Cambridge Scholars Publishing,
Newcastle upon Tyne
Cohen ED, Zinaich S (2013) Philosophy, counseling, and psychotherapy. Cam-
bridge Scholars Publishing, Newcastle upon Tyne
Cohen ED, Piozzini B, Bapat C et al. (2024) A randomized, controlled, preliminary
study to assess the efficacy of logic-based therapy in reducing anxiety and/or
depression in family caregivers. J Ratio Emot Cogn Behav Ther 42:582–609.
https://doi.org/10.1007/s10942-023-00532-z
Cooper M, McLeod J (2010) Pluralistic counselling and psychotherapy. Sage
Daniel MF, Auriac E (2011) Philosophy, critical thinking and philosophy for
children. Educ Philos Theory 43(5):415–435. https://doi.org/10.1111/j.1469-
5812.2008.00483.x
Demszky D, Yang DY, Yeager D et al. (2023) Using large language models in
psychology. Nat Rev Psychol 2(11):688–701. https://doi.org/10.1038/s44159-
023-00241-5
Devlin J, Chang MW, Lee K, Toutanova K (2019, June) Bert: pre-training of deep
bidirectional transformers for language understanding. In: Proceedings of the
2019 Conference of the North American Chapter of the Association for
Computational Linguistics: Human Language Technologies, volume 1 (long
and short papers). ACL, pp. 4171–4186
Ding XJ, Fu SR, Jiao CC, Yu F (2024a) Chinese philosophical practice toward self-
cultivation: integrating Confucian wisdom into philosophical counseling.
Religions 15(1):69. https://doi.org/10.3390/rel15010069
Ding XJ, Harteloh P, Pan TQ, Yu F (2024b) The practical turn in philosophy: a
revival of the ancient art of living through modern philosophical practice.
Metaphilosophy 55(4-5):517–534. https://doi.org/10.1111/meta.12702
Ding XJ, Xie CF, Yu F (2024c) Beyond dissonance: the transformative power of
thought analysis in philosophical practice. Humanit Soc Sci Commun
11(1):1617. https://doi.org/10.1057/s41599-024-04143-6
Ding XJ, Yu F (2024) Philosophical practice and its development in China:
opportunities and challenges. Humanit Soc Sci Commun 11(1):494. https://
doi.org/10.1057/s41599-024-02985-8
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Dreyfus HL (1972) What computers can’t do: a critique of artificial reason. Harper
& Row, New York
Elliott R, Bohart AC, Watson JC, Greenberg LS (2011) Empathy. Psychotherapy
48(1):43–49. https://doi.org/10.1037/a0022187
FatićA, Zagorac I (2016) The methodology of philosophical practice: eclecticism
and/or integrativeness? Philosophia 44(4):1419–1438. https://doi.org/10.
1007/s11406-016-9770-3
Fink MC, Robinson SA, Ertl B (2024) AI-based avatars are changing the way we
learn and teach: benefits and challenges. Front Educ 9:1416307. https://doi.
org/10.3389/feduc.2024.1416307
Fu GH, Zhao Q, Li JQ et al. (2023) Enhancing psychological counseling with large
language model: a multifaceted decision-support system for non-
professionals. https://arxiv.org/abs/2308.15192
Galitsky BA (2024) LLM-based personalized recommendations in health. Pre-
prints. https://doi.org/10.20944/preprints202402.1709.v1
Gao YF, Xiong Y, Gao XY et al. (2024) Retrieval-augmented generation for large
language models: a survey. https://arxiv.org/abs/2312.10997
Gindi S, Pilpel A (2015) Bridging the gap between philosophy and psychotherapy:
an outline for the integration of philosophical counselling into therapeutic
practice. J Humanit Ther 6(2):1–24
Grimes P, Uliana RL (1998) Philosophical midwifery: a new paradigm for
understanding human problems with its validation. Hyparxis Press, Costa
Mesa, CA
Guo Z, Lai A, Thygesen JH, Farrington J, Keen T, Li K (2024) Large language
models for mental health applications: systematic review. JMIR Ment Health
11:e57400. https://doi.org/10.2196/57400
Hadot P (1995) Philosophy as a way of life: spiritual exercises from Socrates to
Foucault. Davidson A (ed.). Translated by Chase M. Wiley-Blackwell, Oxford
Harnad S (1990) The symbol grounding problem. Phys D 42(1-3):335–346. https://
doi.org/10.1016/0167-2789(90)90087-6
Heckmann G (1981) Socratic dialogue: reflections on technique. In: Gronbach K
(ed.), Socratic philosophizing: selected writings. Vandenhoeck & Ruprecht,
Göttingen, pp. 45–59
Hu YH (2024) From theoretical integration to complementary coexistence: an
examination of the inclusiveness of Chinese civilization through the rela-
tionship between Confucianism and Daoism. J Hubei Univ Philos Soc Sci
51(6):83–91. https://doi.org/10.13793/j.cnki.42-1020/c.2024.06.010
Huang Y, Wu RP, He JT, Xiang YP (2024) Evaluating ChatGPT-4.0’s data analytic
proficiency in epidemiological studies: a comparative analysis with SAS, SPSS,
and R. J Glob Health 14:04070. https://doi.org/10.7189/jogh.14.04070
Hume D (1739/2000) A treatise of human nature. Norton DF, Norton MJ (eds.).
Oxford University Press, Oxford
Kandpal N, Deng H, Roberts A, Wallace E, Raffel C (2023) Large language models
struggle to learn long-tail knowledge. In: Proceedings of 40th International
Conference on Machine Learning (ICML). JMLR.org., pp. 15696–15707
Kinderman P (2005) A psychological model of mental disorder. Harv Rev Psychiat
13(4):206–217. https://doi.org/10.1080/10673220500243349
Kjell ON, Kjell K, Schwartz HA (2023) Beyond rating scales: with targeted eva-
luation, language models are poised for psychological assessment. Psychiat
Res 333:115667. https://doi.org/10.1016/j.psychres.2023.115667
Kiriakaki S, Tzovanou M, Sotiriou A, Lampou M, Varsamidou E (2022) Online
counseling: advantages, disadvantages and ethical issues. Homo Virtualis
5(1):32–59. https://doi.org/10.12681/homvir.30316
Knapp S, Tjeltveit AC (2005) A review and critical analysis of philosophical
counseling. Prof Psychol-Res Pr 36(5):558–565. https://doi.org/10.1037/0735-
7028.36.5.558
Kreimer R, Primero G (2017) The future of philosophical counseling: pseu-
doscience or interdisciplinary field? In: Amir L (ed.). New frontiers in phi-
losophical practice. Cambridge Scholars Publishing, Newcastle upon Tyne,
pp. 144–163
Kumar A, Raghunathan A, Jones R, Ma TY, Liang P (2022) Fine-tuning can distort
pretrained features and underperform out-of-distribution. In: Proceedings of
In International Conference on Learning Representations (ICLR). ICLR, pp.
1–42
Kuhail MA, Alturki N, Thomas J, Alkhalifa AK, Alshardan A (2024) Human-
human vs human-AI therapy: an empirical study. Int J Hum-Comput
Interact: 1–12. https://doi.org/10.1080/10447318.2024.2385001
Lahav R, Tillmanns MV (eds) (1995) Essays on philosophical counseling. Uni-
versity Press of America, Lanham
Lee EE, Torous J, De Choudhury M et al. (2021) Artificial intelligence for mental
health care: clinical applications, barriers, facilitators, and artificial wisdom.
Biol Psychiatry Cogn Neurosci Neuroimaging 6(9):856–864. https://doi.org/
10.1016/j.bpsc.2021.02.001
Levkovich I, Elyoseph Z (2023) Suicide risk assessments through the eyes of
ChatGPT-3.5 versus ChatGPT-4: vignette study. JMIR Ment Health
10:e51232. https://doi.org/10.2196/51232
Liu JY, Huang ZY, Xiao T et al. (2024) SocraticLM: exploring socratic personalized
teaching with large language models. In: The Thirty-eighth Annual
Conference on Neural Information Processing Systems (NIPS). Curran
Associates Inc. https://openreview.net/forum?id=qkoZgJhxsA
Liu JM, Li DH, Cao H, Ren TH, Liao ZY, Wu JM (2023) Chatcounselor: a large
language models for mental health support. https://arxiv.org/abs/2309.15461
Liu PF, Yuan WZ, Fu JL, Jiang ZB, Hayashi H, Neubig G (2023) Pre-train, prompt,
and predict: a systematic survey of prompting methods in natural language
processing. ACM Comput Surv 55(9):1–35. https://doi.org/10.1145/3560815
Louw D (2013) Defining philosophical counselling: an overview. S Afr J Philos
32(1):60–70. https://doi.org/10.1080/02580136.2013.810417
Ma XB, Gong YY, He PC, Zhao H, Duan N (2023) Query rewriting in retrieval-
augmented large language models. In: Proceedings of the 2023 Conference on
Empirical Methods in Natural Language Processing (EMNLP). ACL, pp.
5303–5315
Marinoff L (1999) Plato, not Prozac!: applying philosophy to everyday problems.
HarperCollins, New York
Marinoff L (2002) Philosophical practice. Academic Press, San Diego, CA
Ministry of Education of the People’s Republic of China (2023) Number of regular
students for normal courses in HEIs by discipline. http://www.moe.gov.cn/
jyb_sjzl/moe_560/2021/quanguo/202301/t20230103_1037969.html. Accessed
on 12 May 2024
Na HB (2024) CBT-LLM: a Chinese large language model for cognitive behavioral
therapy-based mental health question answering. In: Proceedings of the 2024
Joint International Conference on Computational Linguistics, Language
Resources and Evaluation (LREC-COLING). ELRA and ICCL, pp. 2930–2940
National Center for Education Statistics Database (2023) Table 322.10. Bachelor’s
degrees conferred by postsecondary institutions, by field of study: selected
academic years, 1970-71 through 2021-22. https://nces.ed.gov/programs/
digest/d21/tables/dt21_322.10.asp. Accessed on 12 May 2024
Nelson L (1949) Socratic method and critical philosophy: selected essays. Dover
Publications, New York, NY
O'Neill R, Wang H (2021) Combatting student alienation: community building in
the academic philosophy café. J Humanit Ther 12(1):7–25. https://doi.org/10.
33252/jht.2021.06.12.1.7
Nutas A (2022) The artificial philosophical counselor: on the possibility of auto-
mating philosophical counseling. Int J Philos Pr 8(1):124–136. https://doi.
org/10.5840/ijpp2022819
Obradovich N, Khalsa SS, Khan WU et al. (2024) Opportunities and risks of large
language models in psychiatry. NPP—Digit Psychiatry Neurosci 2(1):1–8.
https://doi.org/10.1038/s44277-024-00010-z
Oliveira FGD, Belitski M, Kakabadse N, Theodorakopoulos N (2024) Unveiling the
potential of digital human avatars in modern marketing strategies. Int Mark
Rev. https://doi.org/10.1108/IMR-12-2023-0339
Park S, Kulkarni C (2023) Thinking assistants: LLM-based conversational assistants
that help users think by asking rather than answering. https://arxiv.org/abs/
2312.06024
Patel SC, Fan J (2023) Identification and description of emotions by current large
language models. bioRxiv. https://doi.org/10.1101/2023.07.17.549421
Pearl J (1988) Probabilistic reasoning in intelligent systems: networks of plausible
inference. Morgan Kaufmann, San Francisco, CA
Petroni F, Rocktäschel T, Lewis P et al. (2019) Language models as knowledge
bases? In: Proceedings of the 2019 Conference on Empirical Methods in
Natural Language Processing and the 9th International Joint Conference on
Natural Language Processing (EMNLP-IJCNLP). ACL, pp. 2463–2473
Rabani ST, Khanday AMUD, Khan QR, Hajam UA, Imran AS, Kastrati Z (2023)
Detecting suicidality on social media: machine learning at rescue. Egypt
Inform J 24(2):291–302. https://doi.org/10.1016/j.eij.2023.04.003
Radford A, Narasimhan K, Salimans T, Sutskever I (2018) Improving language
understanding by generative pre-training. https://openai.com/research/
language-unsupervised
Raile P (2024) The usefulness of ChatGPT for psychotherapists and patients.
Humanit Soc Sci Commun 11:47. https://doi.org/10.1057/s41599-023-02567-0
Repetti R (2023) A sampling of philosophical counseling frameworks, tools, and
practices. Interdiscip Res Couns Ethics Philos 3(8):136–195. https://doi.org/
10.59209/ircep.v3i8.60
Ringle M (2019) Artificial intelligence and semantic theory. In: Simon TW, Scholes
RJ (eds). Language, mind, and brain. Psychology Press, New York, pp. 45–63
Rubin M, Arnon H, Huppert JD, Perry A (2024) Considering the role of human
empathy in AI-driven therapy. JMIR Ment Health 11:e56529. https://doi.org/
10.2196/56529
Savage P (1997) Philosophical counselling. Nurs Ethics 4(1):39–48
Searle JR (1980) Minds, brains, and programs. Behav Brain Sci 3(3):417–424.
https://doi.org/10.1017/S0140525X00005756
Schaaff K, Reinig C, Schlippe T (2023) Exploring ChatGPT’s empathic abilities. In:
2023 11th International Conference on Affective Computing and Intelligent
Interaction (ACII), pp. 1–8. https://doi.org/10.1109/ACII59096.2023.
10388208
Scheler M (1913/1970) The nature of sympathy. Translated by P. Heath. Archon
Books, Hamden, CT
ARTICLE HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7
14 HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Schuster SC (1999) Philosophy practice: an alternative to counseling and psy-
chotherapy. Praeger, Westport & London
Schwitzgebel E, Schwitzgebel D, Strasser A (2023) Creating a large language model of
a philosopher. Mind Lang 39(2):237–259. https://doi.org/10.1111/mila.12466
Shapiro L (2010) Embodied cognition. Routledge, New York, NY
Sharma A, Lin IW, Miner AS, Atkins DC, Althoff T (2023) Human–AI colla-
boration enables more empathic conversations in text-based peer-to-peer
mental health support. Nat Mach Intell 5(1):46–57. https://doi.org/10.1038/
s42256-022-00593-2
Shteynberg G, Halpern J, Sadovnik A et al. (2024) Does it matter if empathic AI has
no empathy? Nat Mach Intell 6:496–497. https://doi.org/10.1038/s42256-024-
00841-7
Sivil R, Clare J (2018) Towards a taxonomy of philosophical counselling. S Afr J
Philos 37(2):131–142. https://doi.org/10.1080/02580136.2018.1432528
Slote M (2007) The ethics of care and empathy. Routledge, London
Smithson R, Zweber A (2024) Reviving the philosophical dialogue with large
language models. Teach Philos. https://doi.org/10.5840/teachphil2024424196
Tashakkori A, Creswell JW (2007) The new era of mixed methods. J Mix Methods
Res 1(1):3–7. https://doi.org/10.1177/2345678906293042
Touvron H, Lavril T, Izacard G et al. (2023a) Llama: open and efficient foundation
language models. https://arxiv.org/abs/2302.13971
Touvron H, Martin L, Stone K et al. (2023b) Llama 2: open foundation and fine-
tuned chat models. https://arxiv.org/abs/2307.09288
Turkle S (2011) Alone together: why we expect more from technology and less
from each other. Basic Books, New York
Turpin M, Michael J, Perez E, Bowman SR (2023) Language models don’t always
say what they think: unfaithful explanations in chain-of-thought prompting.
In: Proceedings of the 37th International Conference on Neural Information
Processing Systems (NIPS). Curran Associates Inc., pp. 74952–74965
Vaswani A, Shazeer N, Parmar N et al. (2017) Attention is all you need. In:
Proceedings of the 31st International Conference on Neural Information
Processing Systems (NIPS). Curran Associates Inc., pp. 6000–6010
Varela FJ, Thompson E, Rosch E (1991) The embodied mind: cognitive science and
human experience. MIT Press, Cambridge, MA
Wankhade M, Rao ACS, Kulkarni C (2022) A survey on sentiment analysis
methods, applications, and challenges. Artif Intell Rev 55(7):5731–5780.
https://doi.org/10.1007/s10462-022-10144-1
Wang M, Vu TT, Wang Y, Shareghi E, Haffari G (2024) Conversational simulMT:
efficient simultaneous translation with large language models. https://arxiv.
org/abs/2402.10552
Wei X, Li Q (2013) The Confucian value of harmony and its influence on Chinese
social interaction. Cross-Cult Commun 9(1):60–66
Wei J, Wang XZ, Schuurmans D et al. (2022) Chain-of-thought prompting elicits
reasoning in large language models. In: Proceedings of the 36th International
Conference on Neural Information Processing Systems (NIPS). Curran
Associates Inc., pp. 24824–24837
Wilson M (2002) Six views of embodied cognition. Psychonom Bullet Rev
9:625–636. https://doi.org/10.3758/BF03196322
Wu XD, Duan R, Ni JB (2024) Unveiling security, privacy, and ethical concerns of
ChatGPT. J Inf Intell 2(2):102–115. https://doi.org/10.1016/j.jiixd.2023.10.
007
Xu XH, Yao BS, Dong YZ et al. (2024) Mental-LLM: leveraging large language
models for mental health prediction via online text data. Proc ACM Interact
Mob Wearable Ubiquit Technol (IMWUT) 8(1):1–32
Yao Y, Duan JH, Xu KD et al. (2024) A survey on large language model (LLM)
security and privacy: the good, the bad, and the ugly. High-Confid Comput
4(2):100211. https://doi.org/10.1016/j.hcc.2024.100211
Zhang B, Haddow B, Birch A (2023) Prompting large language model for machine
translation: a case study. In: Proceedings of 40th International Conference on
Machine Learning (ICML). JMLR.org., pp. 41092–41110
Zhang J, Yan B (2012) The enlightenment of Confucian traditional values on
localized psychological counseling and therapy. Acad Exch 28(4):14–17
Zhao H, Chen H, Yang F et al. (2024) Explainability for large language models: a
survey. ACM T Intel Syst Tec 15(2):1–38. https://doi.org/10.1145/3639372
Zhou SG, Zhang XY (2018) Selecting the best treatment: client-treatment match.
Adv Psychol Sci 26(2):294–305
Zhu YT, Yuan HY, Wang ST et al. (2023) Large language models for information
retrieval: a survey. https://arxiv.org/abs/2308.07107
Acknowledgements
This work was supported by the Natural Science Foundation of Hubei Province (Grant
No. 2023AFB815), the MOE (Ministry of Education in China) Project of Humanities and
Social Sciences (Grant No. 19YJC720006), and the National Social Science Foundation of
China (Grant No. 20FZXB047).
Author contributions
BC and WZ: conceptualization, methodology, writing—original draft. LZ and XD:
conceptualization, resources, supervision, writing—review and editing. All authors have
given their final approval for the version to be published. All authors have agreed to take
responsibility for all aspects of the work to ensure that questions about the accuracy or
integrity of any part of the work are properly investigated and resolved.
Competing interests
Xiaojun Ding was a member of the Editorial Board of this journal at the time of
acceptance for publication. The manuscript was assessed in line with the journal’s
standard editorial processes.
Ethical approval
It does not apply to this article as it does not contain any studies with human
participants.
Informed consent
It does not apply to this article as it does not contain any studies with human
participants.
Additional information
Correspondence and requests for materials should be addressed to Liang Zhao or
Xiaojun Ding.
Reprints and permission information is available at http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons licence, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2025
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | https://doi.org/10.1057/s41599-025-04657-7 ARTICLE
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS | (2025) 12:335 | https://doi.org/10.1057/s41599-025-04657-7 15
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com