Exploring Societal Concerns and Perceptions of AI: A Thematic Analysis through the Lens of Problem-Seeking

Naomi Omeonga wa Kayembe

REVIEW 4 major objections 6 minor 6 references

Exploring Societal Concerns and Perceptions of AI: A Thematic Analysis through the Lens of Problem-Seeking

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper argues that human intelligence cannot be separated from the goals it sets, and that AI, lacking embodied experience, cannot set goals at all.

desk verdict A genuinely interesting conceptual distinction between problem-seeking and problem-solving, but the YouTube study does not empirically test it, and the missing appendices block verification. read the letter →

arxiv 2505.23930 v1 pith:6D54PVY2 submitted 2025-05-29 cs.CY cs.AI

Naomi Omeonga wa Kayembe This is my paper

classification cs.CYcs.AI

keywords problem-seekingproblem-solvingembodiedcognitionorthogonalitythesisAIethicssocietalperceptionsofthematicanalysisYouTubemetadata

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to establish that human intelligence has a distinct component—problem-seeking, the embodied and emotionally grounded act of identifying and setting goals—that current AI lacks. It argues that because humans tie goal-setting to goal-pursuit through bodily experience, the orthogonality thesis, which treats intelligence as separable from goals, describes machines but not people. To test and illustrate the framework, the author analyzes the titles and descriptions of 157 YouTube videos about AI and finds a dual public mood of excitement and anxiety across themes like privacy, job displacement, misinformation, and ethics. If the framework is correct, AI cannot independently decide what to work on, so responsible AI means keeping humans in the goal-setting role and cultivating emotional and digital literacy.

What carries the argument

The load-bearing idea is the problem-seeking/problem-solving distinction, with problem-seeking defined as the embodied, emotionally grounded identification of terminal goals and problem-solving as the execution of strategies toward preset ends. This distinction carries the argument by supplying a criterion for what AI lacks: the paper uses it to reinterpret the embodied-cognition hypotheses of Conceptualization, Replacement, and Constitution, arguing that AI can only offload problem-solving tasks while human goal-setting depends on bodily, non-representational perception and feedback loops. The orthogonality thesis—intelligence separable from goals—serves as the target the paper rejects for human cognition and accepts only for AI. On the empirical side, the machinery is a mixed-methods thematic analysis following the six phases of Braun and Clarke's method, with keyword frequencies and a relevance index built from views, likes, comments, and subscriber counts ranking the 157 videos.

What would settle it

Place a purely computational agent—one with no body, no sensory-motor interaction, and no pre-specified terminal goal—in a novel environment; if it begins to pursue a goal that its designers never programmed and that was never present in its training data, then the claim that AI lacks intrinsic problem-seeking is false.

Watch

Extended reading notes

Core claim

The central claim is that human intelligence is a two-part process: problem-seeking, which grows out of bodily needs, emotions, and context and decides which goals matter, and problem-solving, which devises and executes strategies to reach those goals. The paper contends that these parts are interwoven in humans—what we experience changes what we want, and what we want changes how we think—so the orthogonality thesis, which holds that intelligence is independent of goal content, does not apply to human cognition. Artificial intelligence, by contrast, is described as purely a problem-solver: it optimizes efficiently within goals set externally by designers, because it lacks the embodied, non-representational, bottom-up experience that generates goals. The empirical component, a thematic analysis of 157 YouTube videos, is presented as showing that public discourse about AI is already animated by this gap—people marvel at AI's problem-solving while worrying about privacy, displacement, misuse, and loss of control—and the paper concludes that AI should be framed and regulated as a tool that augments human intelligence rather than a substitute for it.

Load-bearing premise

The paper assumes that the titles and descriptions of 157 popular English-language YouTube videos about AI are a faithful stand-in for broader societal perceptions of AI; if that proxy is unrepresentative, the thematic conclusions about society do not follow.

Editorial extensions

If this is right

Because AI cannot set its own goals, aligning AI with human values is not an optional add-on but the defining requirement of its use.
Public and policy discourse should treat AI as an augmenting tool, keeping goal-setting and value choices in human hands.
Strengthening emotional and digital literacy becomes a practical intervention, since the paper finds public discourse focuses on surface fears while neglecting self-regulation and wisdom.
The orthogonality thesis should not be imported into reasoning about human cognition; comparisons between human and AI intelligence must separate problem-seeking from problem-solving.
Science communication about AI should emphasize accessible, practical demonstrations, since the data suggest technical and speculative content attracts only a niche audience.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural test of the framework would be to run a purely non-embodied agent in a novel environment and check whether it can generate a terminal goal never given by its designers; if it can, the human/AI boundary the paper draws would blur.
The paper's own empirical strategy could be extended to full video transcripts and comment sections, which might reveal that the identified themes change or split when more than titles and descriptions are examined.
If problem-seeking is a skill that can be trained through education, the practical implication is that the human side of the gap can be strengthened—cultivating curiosity and goal formulation may matter as much as improving AI.
The framework implies a division of labor for human-AI collaboration—humans own the 'what' and AI owns the 'how'—which could be tested by comparing outcomes of teams that observe this division against teams that let AI propose goals.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The thing worth knowing about this paper is that the core idea—problem-seeking as a named, embodied counterpart to problem-solving—is actually interesting and not, as far as I can tell, already in the cited literature. The synthesis with embodied cognition and the orthogonality thesis is thoughtfully done, and the writing is clear. The author also deserves credit for stating limitations openly: metadata-only, English-only, self-selected creators, and the appendices being omitted from the preprint. That honesty is real.

The soft spots are proportionate to how much weight the paper puts on them. The central claim—that AI lacks problem-seeking because human goal-setting is embodied—is philosophically argued, not empirically demonstrated. The YouTube themes (privacy, jobs, ethics, etc.) show that people discuss AI's impacts, but they do not show whether human goal-setting is embodied or whether AI's goals are merely delegated. In fact, the paper's own Table 2 says AI adheres more closely to the orthogonality thesis, so the data don't discriminate between the proposed framework and the standard view. The leap from thematic discourse to refuting orthogonality is a conceptual move, not a result. On top of that, the thematic coding has no inter-rater reliability, the keyword correlations are reported without significance tests or effect sizes, the Relevance Index weights are free parameters, and the omitted appendices mean no one can check the theme assignments or keyword lists. The novelty claim also would need a systematic search to back up, not just an assertion.

None of this kills the paper, but it does mean the empirical part should be treated as illustration, not evidence. The author should either release the data and coding materials or explicitly reframe this as a conceptual contribution with a small exploratory case study.

Who should read it: people working on AI ethics, human-AI interaction, and cognitive science who want a fresh label for something that has been discussed in pieces. It could be a useful provocation for a reading group.

My recommendation: send it to peer review, but only with the expectation of major revision. The reviewer should demand the appendices, a reframed empirical section, and a more careful separation of conceptual argument from data interpretation.

Referee Report

4 major / 6 minor

Summary. The paper introduces a conceptual distinction between problem-seeking (the embodied, emotionally grounded process of identifying and setting goals) and problem-solving (executing strategies toward predefined goals), argues that human intelligence non-orthogonally intertwines the two while AI lacks the former, and reports a thematic analysis of metadata from 157 YouTube videos about AI as an empirical exploration. The analysis identifies eleven themes—including privacy, job displacement, misinformation, optimism, and ethical concerns—and offers descriptive keyword and engagement patterns. The paper then uses the problem-seeking lens to critique the orthogonality thesis, to argue that AI cannot independently set goals, and to advocate for emotional and digital literacy and human-centered AI alignment.

Significance. If the conceptual framework were well-supported, it would offer an accessible reframing of human-AI differences and a corrective to purely computational accounts of intelligence. The paper's strengths are its clearly defined conceptual pair, a concrete corpus of 157 YouTube videos, and a detailed thematic taxonomy that could inform public-engagement research. However, the empirical component does not test the core cognitive claim: the identified themes concern AI's impacts and risks and are fully compatible with the orthogonality thesis the paper rejects. The absence of the appendices, the lack of inter-rater reliability, and the unreported inferential statistics make the quantitative results non-verifiable. The paper is best read as a conceptual position paper with an exploratory descriptive study; in its current form the evidence does not license the conclusions drawn.

major comments (4)

[Results/Discussion; Table 2] The central empirical claim—that public themes about AI support the view that humans possess embodied problem-seeking while AI does not—is not tested by the study. The themes identified (privacy, job displacement, misinformation, etc.) concern AI's impacts and risks; they are equally compatible with the standard orthogonality thesis, under which AI has human-external goals and humans deliberate about those goals. Table 2 itself states that AI 'adheres more closely to the orthogonality thesis' and that its 'goals set externally by humans,' which is precisely the position the paper claims to refute. To support the framework, the paper would need a design that directly measures goal-setting behavior, its embodiment, or at minimum a coding scheme that distinguishes delegated goals from autonomous goal generation. As it stands, the YouTube analysis is a post hoc interpretation, not an empirical test.
[Methods – Data preparation and Qualitative analysis; Results] The appendices that would allow independent verification—raw metadata, keyword files, reconstructed titles and descriptions, theme assignment, and Relevance Index details—are omitted: 'Due to file size and format constraints, appendices have not been included in this preprint.' In addition, the many correlational statements in the Results (e.g., 'positive correlation with view count, like count, and comment count') are reported without correlation coefficients, p-values, confidence intervals, or a description of the correlation procedure. A revision should make the appendices available and either provide full inferential statistics or remove the correlational/causal language and present only descriptive patterns.
[Methods – Qualitative analysis] The thematic analysis is conducted by a single coder with no inter-rater reliability, no audit trail, and no discussion of researcher positionality. Because the analysis is based on titles and descriptions rather than full video content, and because the coding is inductive, the risk of selective theme construction is substantial. The paper should report reliability statistics (e.g., Cohen's kappa with a second coder) or explicitly delimit the qualitative results as exploratory and hypothesis-generating.
[Methods – Inclusion Criteria and Procedure] The Relevance Index is a custom algorithm with unspecified free parameters: the weights assigned to view count, engagement (likes/comments), and subscriber count are not given, nor are the threshold values used to select the top 150–200 videos. The inclusion criteria also mix objective filters (duration, language, publication date) with subjective ones ('content focus', 'irrelevant sources') that are not operationalized. Without the full RI formula and decision rules, the sample is not reproducible, which is load-bearing because sample selection determines the themes and keyword frequencies reported.

minor comments (6)

[Methods – Limitations] The paper acknowledges the reliance on YouTube metadata, the English-language restriction, and the omission of non-English-speaking communities, but it does not address the more fundamental representativeness issue that YouTube creators are self-selected and not a probability sample of the general public; a sentence acknowledging this would be appropriate.
[References] Several reference entries are duplicated or inconsistent: the two Carello and Turvey (2005) entries have different titles, Gollwitzer and Sheeran (2006) appears twice, Müller and Cannon (2022) appears as two different papers, Totschnig (2020) appears twice with different titles, and Lemaire (2024) appears twice with different titles; these should be consolidated and corrected.
[Gap in knowledge and research questions] The claim that 'no previous research' has proposed a framework structured around problem-seeking is difficult to evaluate because the paper does not discuss adjacent constructs such as curiosity-driven exploration, active inference, or goal generation in AI safety; a revision should position the concept against these literatures.
[Table 1] Some keyword entries in Table 1 appear to be artifacts of the extraction process rather than meaningful terms, such as 'smartphone-company' and 'Xfounder'; these should be cleaned, merged with their constituent terms, or explained in a note.
[Results – Thematic analysis] The thematic descriptions cite video numbers (e.g., videos 2, 4, 5) but provide no illustrative quotes from the titles or descriptions; adding short example excerpts would strengthen the reader's ability to assess the coding.
[Methods – Ethics] The statement that 'Ethical approval was obtained from the relevant university ethics committee' should identify the committee and provide a protocol reference, as is standard for empirical studies involving human-derived data.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the conceptual problem-seeking framework is not fitted to, and does not reduce to, the YouTube thematic analysis.

full rationale

The paper's central claim—that human intelligence integrates problem-seeking with problem-solving and that AI lacks intrinsic problem-seeking—is grounded in a conceptual argument built from embodied cognition literature (Shapiro, Varela, Gibson, Damasio) and from a stipulated definition of problem-seeking as an embodied, emotionally grounded process. The YouTube metadata study is exploratory and descriptive: it identifies themes (privacy, job displacement, misinformation, optimism) via standard thematic analysis, and those themes are not derived by construction from the problem-seeking definition, nor is the framework fitted to any parameter from the dataset. No prediction is generated from a fitted quantity, and there are no self-citations carrying the argument. The interpretive passages that say the findings 'support the ontological differentiation' are rhetorical readings rather than formal reductions: the themes could also be compatible with the orthogonality view, as the paper's own Table 2 acknowledges. The omitted appendices and the acknowledged limitations of using YouTube metadata are transparency and construct-validity concerns, not circularity. The derivation chain is therefore self-contained at the conceptual level, with the empirical portion serving as illustration rather than as a load-bearing proof of the framework.

Assumptions & free parameters 2 free parameters · 4 assumptions · 1 invented entities

The central claim rests on the new construct of problem-seeking, which is asserted rather than derived, and on assumptions about the representativeness of YouTube metadata and the validity of embodied cognition theories. No independent falsifiable prediction is offered for the framework.

free parameters (2)

Relevance Index (RI) weights
The RI is a weighted sum of view count, likes, comments, and subscriber count, but the weights are not specified. The ranking and final sample selection are therefore not reproducible.
Inclusion thresholds = 150-200 videos, 5+ minutes, last 3 years, English, certain countries
These thresholds are arbitrary choices that shape the dataset; they are not derived from theory or prior work.

assumptions (4)

domain assumption YouTube metadata is representative of societal perceptions of AI
Invoked in Methods; the study treats video titles and descriptions as public opinion data.
domain assumption Thematic analysis per Braun and Clarke (2012) is appropriate for this data
Used as the analytical method; its validity is assumed.
domain assumption The orthogonality thesis is accurately characterized by Bostrom (2014) and Armstrong (2013)
The critique in Discussion relies on this characterization.
domain assumption Embodied cognition theories (Shapiro, Varela) provide a correct account of human cognition
The framework rests on these theories, which are themselves contested.

invented entities (1)

Problem-seeking
purpose: A proposed dimension of intelligence that identifies and sets goals, grounded in embodied affect and experience.
It is defined in the paper but not operationalized or measured independently; it is a conceptual construct used to interpret existing data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Exploring Societal Concerns and Perceptions of AI: A Thematic Analysis through the Lens of Problem-Seeking." pith.science (2026). https://pith.science/paper/6D54PVY2

@misc{pith2026250523930,
  author       = {Pith},
  title        = {Pith review of: Exploring Societal Concerns and Perceptions of AI: A Thematic Analysis through the Lens of Problem-Seeking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6D54PVY2}},
  note         = {Machine review of arXiv:2505.23930}
}

read the original abstract

This study introduces a novel conceptual framework distinguishing problem-seeking from problem-solving to clarify the unique features of human intelligence in contrast to AI. Problem-seeking refers to the embodied, emotionally grounded process by which humans identify and set goals, while problem-solving denotes the execution of strategies aimed at achieving such predefined objectives. The framework emphasizes that while AI excels at efficiency and optimization, it lacks the orientation derived from experiential grounding and the embodiment flexibility intrinsic to human cognition. To empirically explore this distinction, the research analyzes metadata from 157 YouTube videos discussing AI. Conducting a thematic analysis combining qualitative insights with keyword-based quantitative metrics, this mixed-methods approach uncovers recurring themes in public discourse, including privacy, job displacement, misinformation, optimism, and ethical concerns. The results reveal a dual sentiment: public fascination with AI's capabilities coexists with anxiety and skepticism about its societal implications. The discussion critiques the orthogonality thesis, which posits that intelligence is separable from goal content, and instead argues that human intelligence integrates goal-setting and goal-pursuit. It underscores the centrality of embodied cognition in human reasoning and highlights how AI's limitations come from its current reliance on computational processing. The study advocates for enhancing emotional and digital literacy to foster responsible AI engagement. It calls for reframing public discourse to recognize AI as a tool that augments -- rather than replaces -- human intelligence. By positioning problem seeking at the core of cognition and as a critical dimension of intelligence, this research offers new perspectives on ethically aligned and human-centered AI development.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

6 extracted references · 5 canonical work pages

[1]

black box

icter requirements on high-risk AI systems used in critical areas such as healthcare, transportation, and law enforcement. The AI Act also addresses issues such as transparency, requiring AI systems to be explainable to users and mandating disclosure when individuals are interacting with AI rather than humans. This focus on transparency reflects broader s...

work page 2016
[3]

Video 1” or “Video 2

How can the concept of problem-seeking reframe the identified societal concerns about AI? 14 By addressing these research questions, our investigation contributes to a deeper understanding of societal perceptions of AI, and the importance of incorporating public attitudes into the development and deployment of AI technologies. This approach not only provi...

work page 2012
[4]

Frontal Lobe

Kim, T. W., & Scheller-Wolf, A. (2019). Technological unemployment, meaning in life, purpose of business, and the future of stakeholders. Journal of Business Ethics, 160(2), 319–337. https://doi.org/10.1007/s10551-019-04205-9 Lakoff, G., & Johnson, M. (1980). Metaphors We Live By. University of Chicago Press. Lakoff, G., & Johnson, M. (1999). Philosophy I...

arXiv 2019
[157]

turns crazier than I thought,

This theme addresses user-friendly AI tools and accessibility, broadening AI usage in consumer products, and the integration of AI in everyday devices and user base expansion. The sentiments here are optimism and hope, and curiosity and enthusiasm. Qualitative refinement with quantitative analysis The quantitative analysis focused on the frequency of keyw...

work page 1988
[1597]

D., & Russell, S

Catalyzing Next-generation Artificial Intelligence through NeuroAI Zhu, W., Hadfield-Menell, D., Dragan, A. D., & Russell, S. J. (2018). Value alignment, rationality, and intelligence. In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society (pp. 421-426). *Due to file size and format constraints, appendices have not been included in this...

work page 2018
[2005]

intelligent

preventing “intelligent” machines from replicating human problem-seeking abilities. This distinction ties into the differentiation between embodied cognition and embodied action (Carello & Turvey, 2005). While specific cognitive tasks can indeed be offloaded to non-biological systems (embodied cognition, as implied by the Constitution Hypothesis), such ta...

work page arXiv 2005

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

black box

icter requirements on high-risk AI systems used in critical areas such as healthcare, transportation, and law enforcement. The AI Act also addresses issues such as transparency, requiring AI systems to be explainable to users and mandating disclosure when individuals are interacting with AI rather than humans. This focus on transparency reflects broader s...

work page 2016

[2] [3]

Video 1” or “Video 2

How can the concept of problem-seeking reframe the identified societal concerns about AI? 14 By addressing these research questions, our investigation contributes to a deeper understanding of societal perceptions of AI, and the importance of incorporating public attitudes into the development and deployment of AI technologies. This approach not only provi...

work page 2012

[3] [4]

Frontal Lobe

Kim, T. W., & Scheller-Wolf, A. (2019). Technological unemployment, meaning in life, purpose of business, and the future of stakeholders. Journal of Business Ethics, 160(2), 319–337. https://doi.org/10.1007/s10551-019-04205-9 Lakoff, G., & Johnson, M. (1980). Metaphors We Live By. University of Chicago Press. Lakoff, G., & Johnson, M. (1999). Philosophy I...

arXiv 2019

[4] [157]

turns crazier than I thought,

This theme addresses user-friendly AI tools and accessibility, broadening AI usage in consumer products, and the integration of AI in everyday devices and user base expansion. The sentiments here are optimism and hope, and curiosity and enthusiasm. Qualitative refinement with quantitative analysis The quantitative analysis focused on the frequency of keyw...

work page 1988

[5] [1597]

D., & Russell, S

Catalyzing Next-generation Artificial Intelligence through NeuroAI Zhu, W., Hadfield-Menell, D., Dragan, A. D., & Russell, S. J. (2018). Value alignment, rationality, and intelligence. In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society (pp. 421-426). *Due to file size and format constraints, appendices have not been included in this...

work page 2018

[6] [2005]

intelligent

preventing “intelligent” machines from replicating human problem-seeking abilities. This distinction ties into the differentiation between embodied cognition and embodied action (Carello & Turvey, 2005). While specific cognitive tasks can indeed be offloaded to non-biological systems (embodied cognition, as implied by the Constitution Hypothesis), such ta...

work page arXiv 2005