Pith. sign in

REVIEW 4 major objections 6 minor 284 references

Ethical rules for financial mental-health AI can be written as logic and checked so violations are impossible inside the model.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-31 19:24 UTC pith:FGTKRD5E

load-bearing objection Thin but honest early encoding: deontic constraints plus Z3 checks that mostly restate the rules; useful domain sketch, not yet a verification result that does real work. the 4 major comments →

arxiv 2607.24275 v1 pith:FGTKRD5E submitted 2026-07-27 cs.LO cs.AIcs.CY

A Computational Ethical Framework for Financial Digital Phenotyping for Mental Health

Oluwadara Adedeji , Michael Mayowa Farayola , Jeff Brozena , Irina Tal , Regina Connolly , Mark Matthews This is my paper
classification cs.LO cs.AIcs.CY
keywords computational ethicsdigital phenotypingfinancial datadeontic temporal logicSMT verificationinformed consentfairnessmental health
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

High-level ethics rules for AI rarely become something a running system can check. This paper targets that gap for digital phenotyping that uses financial behaviour to track mental health. It turns requirements such as valid consent, withdrawal, data minimisation, and non-discrimination into deontic temporal logic constraints, and defines a conceptual ethical oversight agent that only counts as ethical if every system it supervises satisfies those constraints. Using the Z3 solver, the authors show the constraint set is consistent and that crafted violation scenarios are unsatisfiable, so the formal model rules those breaches out. The point for a sympathetic reader is a path from static compliance paperwork toward continuous, machine-checkable ethical guarantees in a high-risk health setting, while still calling for human oversight where capacity and fairness need judgement.

Core claim

Within a formal model of financial digital phenotyping, key ethical properties—informed consent, post-withdrawal invalidation, ban on collection or use without valid consent, ban on unnecessary identifying fields, and responsibility via fairness and non-discrimination—can be encoded as deontic temporal logic constraints plus an ethical-agent condition; Z3 finds the set satisfiable and every tested violation scenario unsatisfiable, so those breaches are ruled out inside the model.

What carries the argument

Deontic temporal logic constraints (obligations and prohibitions over time) on predicates for consent, collection, use, fairness, and supervision, compiled for Z3 counterexample checks; the ethical agent is ethical only if every supervised system satisfies the full constraint bundle.

Load-bearing premise

That hard-to-judge notions like capacity, voluntariness, fairness, and discrimination can be treated as clear yes-or-no facts the system can check, so passing the logic equals real ethical compliance.

What would settle it

Run the same Z3-backed checks on a real financial digital-phenotyping pipeline with logged consent, withdrawal, field collection, and outcome events: if a live trace that clearly violates consent or fairness still satisfies the formal constraints, or a compliant trace is rejected, the central claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Consent, withdrawal, data minimisation, and fairness duties from high-level rules (including EU AI Act data-governance and bias language) can be turned into machine-checkable constraints rather than static documents alone.
  • An oversight agent can be defined so it is ethical only when every system it supervises meets the full constraint set, supporting hybrid human–AI monitoring.
  • Counterexample-style verification can reject invalid-consent, post-withdrawal use, unnecessary identifying fields, and unfair outcomes inside the formal model before deployment sandboxes.
  • Financial behavioural signals for mental health become a concrete testbed for continuous auditable ethical checking in high-risk AI.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If predicates for capacity and fairness stay human-labelled, the solver mostly audits consistency of labels rather than discovering ethics in the wild—neurosymbolic bridges the paper hints at would be the next engineering step.
  • The same deontic bundle could be reused for other passive phenotyping streams (location, keyboard, wearables) once domain predicates are swapped in.
  • Regulators could treat unsat violation queries as auditable evidence artefacts alongside DPIAs, not as a replacement for them.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a computational ethics framework for financial digital phenotyping in mental health. Ethical requirements—informed consent (validity conditions, withdrawal), data minimisation for identifying financial fields, and responsibility/fairness/non-discrimination—are formalised as deontic temporal logic constraints (Eqs. 1–5), together with an "ethical oversight agent" condition (Eq. 6) stating that an ethical agent supervises only systems satisfying the constraints. These constraints are asserted globally in Z3; the authors report (i) global satisfiability (with an acknowledged vacuous model in which supervision, collection, and use are all false), (ii) unsatisfiability of six "counterexample" scenarios each violating one constraint, and (iii) a positive test permitting collection when all conditions hold. The constraints are mapped to Articles 10 and 14 of the EU AI Act, and the implementation is publicly available. The authors frame this as early work toward continuous, machine-verifiable ethical checking.

Significance. The application area is timely and genuinely under-served: financial data as a digital phenotyping modality raises consent, privacy, and fairness issues for which essentially no formal treatment exists, and the attempt to operationalise EU AI Act obligations as checkable constraints is a worthwhile direction. The paper has real strengths: a public, reproducible Z3 implementation; an unusually candid limitations section (context-dependence of capacity/voluntariness, absence of alternative legal bases, socio-technical misuse risks); and honest reporting of the vacuous global model. The predicate inventory and constraint set are a reasonable starting taxonomy. However, the technical contribution as it stands is thin: the verification results establish only the internal consistency of a small set of universally quantified implications and the unsurprising fact that each constraint contradicts its own negation. If the authors can connect the constraints to an actual system model (traces or transition relations) and calibrate the claims, the framework could become a useful foundation for auditable compliance tooling in this domain.

major comments (4)
  1. [§3.2 (Verification), counterexample methodology] The load-bearing evaluation claim — that violations are 'ruled out' — rests on checks that cannot fail by construction. Each counterexample test asserts Eqs. (1)–(5) globally and then asserts a scenario that syntactically contradicts one of them (e.g., the 'data use without consent' scenario asserts exactly (Collect ∨ UseData) ∧ ¬ConsentValid, which Eq. (3) forbids). Unsat follows by propositional logic alone; no theory reasoning, deontic structure, or temporal reasoning (G over t′ > t) does any work in these results. There is no model of a system being verified — no transition relation, no execution traces, no implementation checked against the constraints. The paper should either (a) reframe the claims to state precisely what was verified (consistency of the constraint set and non-over-constraint via the positive test), or (b) add a system-level artefact — e.g., a set of concrete trace
  2. [§3.1.1, Eq. (6) and §3.2] The 'ethical oversight agent ensures that any supervised system satisfies the constraints' claim is never instantiated non-vacuously. Eq. (6) is a one-way implication, Ethical(A) → ∀S(Supervises(A,S) → Constraints(S)); the authors themselves report that the only satisfying global model assigns Supervises and all system behaviours False. So no model is exhibited in which an ethical agent actually supervises a system that actually collects or uses data, and the prose ('oversees the system and ensures...', Abstract and §2) is stronger than the formalisation ('is considered ethical only if'). Please either provide a positive model satisfying Ethical(A) ∧ Supervises(A,S) ∧ Collect(S,f,p,t) ∧ Constraints(S), or weaken the prose to match Eq. (6). As written, the agent contributes no constraint beyond Eqs. (1)–(5) and the verification section gives it no content.
  3. [§3.1, encoding of deontic operators (Eqs. 1–5)] The paper does not state how the deontic operators O and F are encoded in Z3. Given F(φ) ≡ O(¬φ) (§2) and the reported results, the encoding appears to collapse obligations to material implications (F(φ) ↦ ¬φ; O(ψ) ↦ ψ). This has consequences that should be addressed: (i) Eq. (1) as O(ConsentValid → Cap ∧ Info ∧ Vol) is vacuously satisfied whenever consent is invalid — the obligation is discharged precisely in the states where it matters; the natural formalisation is a definitional biconditional/conditional for validity, or an obligation on the antecedent's conditions when consent is claimed. (ii) All constraints are jointly satisfied by a do-nothing system (never collect, never use), which is consistent but means 'compliance' in the model does not imply any ethically functioning system exists. (iii) If O and F carry no logical weight beyond material implication, the 'deontic temporal lo
  4. [§3.1.1 vs. Abstract/§4/§5, claims calibration] Several claims exceed what is demonstrated. (a) 'Continuous' checking (Abstract, Conclusion): nothing in §3.2 is continuous; all checks are one-shot satisfiability queries. Either demonstrate a runtime-monitoring setup (constraints re-checked on event streams) or drop the word. (b) 'Regulatory-ready pathway' (§4): the mapping to Article 10 is asserted at the level of keyword correspondence; a table aligning each Article 10 paragraph to a formalised (or unformalised) requirement would make this claim concrete and would honestly expose the gaps (e.g., data governance record-keeping, representativeness of datasets). (c) 'To our knowledge, the first work to apply computational ethics methods to digital phenotyping' (§5): plausible for this specific modality, but it should be hedged and situated against the adjacent formal-compliance literature (normative multi-agent systems, formal GDPR/AI-A
minor comments (6)
  1. [§2, predicate/domain table] The time domain T is never specified (discrete? dense? bounded?), yet Eq. (2) quantifies over t′ > t and the Z3 encoding must commit to a theory (integers/reals). Please state the domain and how quantified formulas are handled by the solver (quantifier instantiation, any patterns/triggers), since Z3 can return 'unknown' on quantified inputs; report solver version and that no 'unknown' results occurred.
  2. [§2/§3.1.1] Eq. (4) uses the research-purpose variable ρ but the constraints never bind or constrain ρ, and §4 admits the framework does not evaluate whether ρ itself is justified. Consider quantifying ρ explicitly (e.g., 'for the declared purpose ρ₀') in Eqs. (4) and (7) so the formalisation matches the prose.
  3. [§3.1.1(d), Eq. (5)] Fair(S), BiasMitigated(S), and Discriminate(S,p,t) are atomic predicates with no defined semantics, yet §4 correctly notes these require human judgement. A sentence in §3.1 stating that these atoms are assumed to be supplied by an external assessor (and naming who/what that assessor could be) would preempt misreading of the verification as establishing actual fairness.
  4. [Typesetting] Several PDF-extraction artefacts appear in the manuscript: 'V ol' for Vol in the predicate table, 'F φ' and 'G foralways' in §2, 'Dupr ´e' in reference [10], and a missing space in 'system in which' in the Abstract ('phenotyping system in which'). Please proof the final PDF.
  5. [References] The formal-methods/compliance literature is under-cited: beyond Giordano et al. [10], relevant strands include normative multi-agent systems, formalisation of GDPR/data-protection norms, and runtime verification of norm-governed systems. Two or three anchor citations would help position the contribution and support (or qualify) the novelty claim.
  6. [Availability] It is commendable that the code is public. Please pin a commit or release tag in the camera-ready and include a README mapping each reported sat/unsat result in §3.2 to a script, so the table of results is directly reproducible.

Circularity Check

3 steps flagged

Counterexample 'verification' and the ethical-agent claim are largely true by construction: unsat scenarios restate the asserted constraints, and Ethical(A) is defined as supervising only constraint-satisfying systems.

specific steps
  1. self definitional [§3.1.1 Explicit Ethical Agent, Eqs. (6)–(7); cf. Abstract]
    "Explicit Ethical Agent: An agent A is considered ethical only if every system it supervises satisfies all specified ethical constraints ((a)–(d)) Ethical(A)→ ∀S∈ S (Supervises(A, S)→Constraints(S)) Constraints(S)≜Eq.(1)∧Eq.(2)∧Eq.(3)∧Eq.(4)∧Eq.(5). Abstract: 'a conceptual ethical agent that oversees the system and ensures that any supervised system satisfies the specified constraints.'"

    Ethical(A) is defined as the implication that supervised systems satisfy Constraints(S). The claim that the agent 'ensures' compliance is therefore identical to the definition of the predicate Ethical, not a derived or verified enforcement property. With the global sat model assigning Supervises and system behaviours False (Results 3.2), the implication holds vacuously and does not exhibit non-trivial oversight.

  2. self definitional [§3.2 Verification; Eqs. (1)–(5) vs. counterexample scenarios]
    "We evaluated the framework using counterexample-based verification by formulating a violating scenario for each ethical property and submitting it to the solver. These scenarios included invalid informed consent, continued consent following withdrawal, data collection or use without valid consent, unnecessary collection of identifying financial data, violations of responsible AI requirements, and an ethical oversight agent supervising a non-compliant system. In every case, the solver returned unsatisfiable, confirming that the proposed constraints prevent the corresponding ethical violation."

    Each test asserts Constraints (e.g. F((Collect∨UseData)∧¬ConsentValid) from Eq. 3) together with a scenario that asserts the forbidden conjunction. Unsat is φ∧¬φ by construction inside the formal model; Z3 is not checking an independent transition system or external evidence. The evaluation claim that 'violations are ruled out' therefore restates that the authors forbade those patterns, which cannot fail given how the queries are built. The paper notes the global model is vacuous, underscoring that no realistic compliant behaviour is verified.

  3. other [§3.2 global satisfiability; Discussion]
    "The global satisfiability check returned sat, indicating that Eqs. (1)–(6) are logically consistent and that at least one assignment of variables and predicates satisfies all constraints. However, the generated satisfying model is vacuous, as several key system behaviours, including data collection, data use, withdrawal, and supervision, are assigned False by the solver. This is expected because the global satisfiability check verifies logical consistency rather than realistic system behaviour."

    Consistency is obtained by the trivial all-false assignment that never collects, uses, withdraws, or supervises. Framing sat + counterexample-unsat as 'continuous, machine-verifiable ethical checking' and a path beyond static documentation overstates a result that holds only because forbidden atoms are never forced true and violation queries negate the axioms. This is content-thin verification rather than an independent predictive or first-principles derivation.

full rationale

The paper's load-bearing evaluation claim is that Z3 shows the deontic constraints are consistent and that violation scenarios are ruled out, and that a conceptual ethical agent oversees systems to ensure compliance. Both results largely reduce to the inputs by construction rather than to independent content. (1) Each counterexample test asserts the global constraints (Eqs. 1–5) and then a scenario that is the syntactic negation of one constraint (e.g., Collect/UseData with ¬ConsentValid against Eq. 3); unsatisfiability follows from propositional contradiction inside the formal model, not from checking an independent system model, traces, or external normative evidence. The paper itself reports that the only globally satisfying assignment is vacuous (collection, use, withdrawal, and supervision all False), so nothing non-trivial is shown to obey the norms. (2) Eq. 6 defines Ethical(A) via the one-way implication that every supervised system satisfies Constraints(S); the abstract/results claim that the agent 'ensures' supervised systems satisfy the constraints therefore restates the definition of Ethical rather than deriving an enforcement property. There is no self-citation chain, fitted empirical parameter, or renamed external theorem—the circularity is internal and definitional/tautological. The formalisation of consent/fairness norms as deontic formulas still has independent specification content; the circularity is concentrated in what the evaluation is said to establish. Score 6 reflects partial circularity on the central verification and agent claims without collapsing the entire contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 6 axioms · 2 invented entities

The central claim rests on standard deontic/temporal identities, on treating contested ethical concepts as boolean predicates, on identifying a small constraint set with EU AI Act Articles 10/14-style duties, and on equating Z3 unsat of hand-written violation scenarios with ruled-out unethical behaviour inside the model. No numeric fitting; main load is domain and modelling assumptions plus the invented oversight-agent construct.

axioms (6)
  • standard math Prohibition is equivalent to obligation of negation: F(φ) ≡ O(¬φ); temporal operator G expresses always/future invalidation after withdrawal.
    Used throughout §2–3 to compile deontic rules into solver constraints; standard deontic/temporal logic.
  • domain assumption Informed consent validity is captured by Cap(p) ∧ Info(p) ∧ Vol(p) as necessary conditions (Eq. 1), with withdrawal implying G t'>t ¬ConsentValid (Eq. 2).
    From institutional ethics/DPIA and literature; treats capacity, information adequacy, and voluntariness as formal atoms evaluable in the model.
  • domain assumption Data collection/use without valid consent, and collection of identifying fields not Necessary(f,ρ), are forbidden (Eqs. 3–4).
    Data-governance / minimisation norms aligned by authors with EU AI Act Art. 10; necessity relative to an uninterpreted research purpose ρ.
  • domain assumption Responsible(S) requires Fair(S) ∧ BiasMitigated(S) ∧ ∀p,t ¬Discriminate(S,p,t) (Eq. 5).
    Responsibility reduced to three uninterpreted fairness-related atoms; transparency/accountability explicitly omitted.
  • ad hoc to paper Z3 unsatisfiability of encoded violation scenarios constitutes verification that ethical properties are ruled out within the formal model.
    Evaluation design in §3.2; equates solver outcomes on hand-crafted scenarios with ethical verification without external behavioural traces.
  • ad hoc to paper Focus on data collection and modelling stages is sufficient for the claimed framework contribution in current financial phenotyping research.
    Stated in Methods; deployment, retention, and lifecycle are out of scope yet abstract claims continuous governance.
invented entities (2)
  • Conceptual ethical oversight agent A with Ethical(A) ↔ supervises only systems satisfying Constraints(S) no independent evidence
    purpose: Meta-level supervisor that defines ethical agency as universal compliance of supervised phenotyping systems.
    Introduced in Eqs. 6–7 and Discussion (possible LLM implementation); no implemented agent, only a definitional constraint in Z3 scenarios.
  • Predicate suite (ConsentValid, IdentifyingField, Necessary, Fair, BiasMitigated, Discriminate, Responsible, Supervises) no independent evidence
    purpose: Atomic interface between informal ethics/regulation and SMT constraints.
    Paper-defined vocabulary for the case study; semantics left largely open and acknowledged as subjective/contextual.

pith-pipeline@v1.2.0-grok45-kimik3 · 11595 in / 3601 out tokens · 74618 ms · 2026-07-31T19:24:58.283515+00:00 · methodology

0 comments
read the original abstract

Ethical governance of AI-driven systems is often expressed through high-level principles and static documentation, creating a gap between regulatory requirements and system-level verification. This challenge is particularly acute in digital phenotyping, where continuous behavioural data raises concerns around consent, privacy, and fairness. In this paper, we propose a computational ethical framework for AI-driven digital phenotyping system in which ethical requirements are formalised as deontic temporal logic constraints, alongside a conceptual ethical agent that oversees the system and ensures that any supervised system satisfies the specified constraints. Using a case study involving financial data and mental health, we model key ethical properties and verify them using the Z3 Satisfiability Modulo Theories (SMT) solver. Our evaluation shows that the framework is logically consistent and that violations of the specified ethical properties are ruled out within the formal model through counterexample-based verification. This presents early research enabling continuous, machine-verifiable ethical checking, moving beyond retrospective compliance based on static documentation. We discuss limitations, including the need for real-world verification with data, the challenge with subjectivity and contextual sensitivity, the need for human oversight, and outline how such approaches can support the development of digital phenotyping and AI systems with continuous and auditable ethical guarantees.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

284 extracted references · 153 canonical work pages · 12 internal anchors

  1. [1]

    Regulation (

    European Parliament. Regulation (. 2024 , note =

    2024

  2. [2]

    2020 , note =

    Data Mining and Knowledge Discovery , author =. 2020 , note =. doi:10.1007/s10618-020-00701-z , abstract =

    doi:10.1007/s10618-020-00701-z 2020

  3. [3]

    AI and Ethics , author =

    Beyond aggregate fairness: intersectional auditing across the. AI and Ethics , author =. 2026 , keywords =. doi:10.1007/s43681-025-00942-y , abstract =

    work page doi:10.1007/s43681-025-00942-y 2026

  4. [4]

    Temporal deontic action logic for the verification of compliance to norms in

    Giordano, Laura and Martelli, Alberto and Dupré, Daniele Theseider , month = jun, year =. Temporal deontic action logic for the verification of compliance to norms in. Proceedings of the. doi:10.1145/2514601.2514608 , abstract =

    arXiv

  5. [5]

    What makes clinical machine learning fair?

    Hoche, Marine and Mineeva, Olga and Rätsch, Gunnar and Vayena, Effy and Blasimme, Alessandro , month = mar, year =. What makes clinical machine learning fair?. PLOS Digital Health , publisher =. doi:10.1371/journal.pdig.0000728 , abstract =

    work page doi:10.1371/journal.pdig.0000728

  6. [6]

    and Li, Mengyuan and Hahne, Jessica and Wang, Junyi and Ren, Yongqi and Yang, Lei and Zhang, Xin and Liu, Xing and Wang, Xiaomin and Bogdan, Ryan and Carpenter, Brian D

    Gorelik, Aaron J. and Li, Mengyuan and Hahne, Jessica and Wang, Junyi and Ren, Yongqi and Yang, Lei and Zhang, Xin and Liu, Xing and Wang, Xiaomin and Bogdan, Ryan and Carpenter, Brian D. , month = sep, year =. Ethics of. Frontiers in Digital Health , publisher =. doi:10.3389/fdgth.2025.1662642 , abstract =

    arXiv 2025

  7. [7]

    Ethics in

    Milosevic, Zoran , month = oct, year =. Ethics in. 2019. doi:10.1109/EDOC.2019.00022 , abstract =

    arXiv 2019

  8. [9]

    Science and Engineering Ethics , author =

    Ethical,. Science and Engineering Ethics , author =. 2022 , pages =. doi:10.1007/s11948-021-00354-1 , abstract =

    work page doi:10.1007/s11948-021-00354-1 2022

  9. [10]

    2020 , note =

    Data Mining and Knowledge Discovery , author =. 2020 , note =. doi:10.1007/s10618-020-00710-y , abstract =

    doi:10.1007/s10618-020-00710-y 2020

  10. [11]

    and Webb, Geoffrey I

    Dempster, Angus and Schmidt, Daniel F. and Webb, Geoffrey I. , month = mar, year =. doi:10.48550/arXiv.2203.13652 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2203.13652

  11. [12]

    and Webb, Geoffrey I

    Dempster, Angus and Schmidt, Daniel F. and Webb, Geoffrey I. , month = aug, year =. Proceedings of the 27th. doi:10.1145/3447548.3467231 , abstract =

    arXiv

  12. [13]

    Formal specification and verification of user-centric privacy policies for ubiquitous systems , isbn =

    Joshaghani, Rezvan and Black, Stacy and Sherman, Elena and Mehrpouyan, Hoda , month = jun, year =. Formal specification and verification of user-centric privacy policies for ubiquitous systems , isbn =. Proceedings of the 23rd. doi:10.1145/3331076.3331105 , abstract =

    arXiv

  13. [14]

    and Rao, Shrisha , month = jan, year =

    V, Priya T. and Rao, Shrisha , month = jan, year =. Deontic. doi:10.48550/arXiv.2501.05765 , abstract =

    doi:10.48550/arxiv.2501.05765

  14. [15]

    Journal of Technology in Behavioral Science , author =

    Ethical. Journal of Technology in Behavioral Science , author =. 2025 , keywords =. doi:10.1007/s41347-024-00423-9 , abstract =

    work page doi:10.1007/s41347-024-00423-9 2025

  15. [16]

    and Rahwan, Iyad , month = dec, year =

    Kim, Richard and Kleiman-Weiner, Max and Abeliuk, Andrés and Awad, Edmond and Dsouza, Sohan and Tenenbaum, Joshua B. and Rahwan, Iyad , month = dec, year =. A. Proceedings of the 2018. doi:10.1145/3278721.3278770 , abstract =

    arXiv 2018

  16. [17]

    Trends in Cognitive Sciences , author =

    Computational ethics , volume =. Trends in Cognitive Sciences , author =. 2022 , keywords =. doi:10.1016/j.tics.2022.02.009 , abstract =

    work page doi:10.1016/j.tics.2022.02.009 2022

  17. [18]

    AI and Ethics , author =

    Formalizing ethical principles within. AI and Ethics , author =. 2025 , pages =. doi:10.1007/s43681-024-00425-6 , abstract =

    work page doi:10.1007/s43681-024-00425-6 2025

  18. [19]

    AI and Ethics , author =

    Trends and challenges in machine ethics: an updated survey of artificial moral agency implementations , volume =. AI and Ethics , author =. doi:10.1007/s43681-025-00971-7 , abstract =

    work page doi:10.1007/s43681-025-00971-7

  19. [20]

    Mapping the ethical landscape of digital biomarkers:

    Andreoletti, Mattia and Haller, Luana and Vayena, Effy and Blasimme, Alessandro , month = may, year =. Mapping the ethical landscape of digital biomarkers:. PLOS Digital Health , publisher =. doi:10.1371/journal.pdig.0000519 , abstract =

    work page doi:10.1371/journal.pdig.0000519

  20. [22]

    AI and Ethics , author =

    Development of. AI and Ethics , author =. 2025 , keywords =. doi:10.1007/s43681-025-00851-0 , abstract =

    work page doi:10.1007/s43681-025-00851-0 2025

  21. [23]

    Applied Artificial Intelligence , publisher =

    Radanliev, Petar , month = dec, year =. Applied Artificial Intelligence , publisher =. doi:10.1080/08839514.2025.2463722 , abstract =

    arXiv 2025

  22. [24]

    Psychological Medicine , author =

    Ethical decision-making for. Psychological Medicine , author =. 2025 , pages =. doi:10.1017/S0033291725101311 , abstract =

    work page doi:10.1017/s0033291725101311 2025

  23. [25]

    , month = feb, year =

    Hossain, Rakib and Khan, Mahmood Menon and Amin, Lisan Al and Parikh, Dhruv and Afroz, Farhana and Ahmed, Bestoun S. , month = feb, year =. Ethical and. doi:10.48550/arXiv.2603.03341 , abstract =

    work page doi:10.48550/arxiv.2603.03341

  24. [26]

    Philosophy & Technology , author =

    Ethical. Philosophy & Technology , author =. 2021 , keywords =. doi:10.1007/s13347-021-00445-8 , abstract =

    work page doi:10.1007/s13347-021-00445-8 2021

  25. [27]

    Stephenson, Elizabeth and Drysdale, Erik and Erdman, Lauren and Goldenberg, Anna and Zlotnik Shaul, Randi , month = may, year =

    McCradden, Melissa D and Anderson, James A and A. Stephenson, Elizabeth and Drysdale, Erik and Erdman, Lauren and Goldenberg, Anna and Zlotnik Shaul, Randi , month = may, year =. A. The American Journal of Bioethics , publisher =. doi:10.1080/15265161.2021.2013977 , abstract =

    arXiv 2021

  26. [28]

    An End-To-End Machine Learning Pipeline That Ensures Fairness Policies

    Shaikh, Samiulla and Vishwakarma, Harit and Mehta, Sameep and Varshney, Kush R. and Ramamurthy, Karthikeyan Natesan and Wei, Dennis , month = oct, year =. An. doi:10.48550/arXiv.1710.06876 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1710.06876

  27. [29]

    doi:10.48550/arXiv.2601.08878 , abstract =

    Steigerwald, Philipp and Burghardt, Jennifer and Rudolph, Eric and Albrecht, Jens , month = jan, year =. doi:10.48550/arXiv.2601.08878 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2601.08878

  28. [30]

    Balancing the

    Bassil, Katherine , month = feb, year =. Balancing the. The American Journal of Bioethics , publisher =. doi:10.1080/15265161.2023.2296437 , number =

    arXiv 2023

  29. [31]

    and Borbón, Diego and Bezerra, Ana Maria , year =

    Muñoz, José M. and Borbón, Diego and Bezerra, Ana Maria , year =. Computational psychiatry and digital phenotyping:. Developments in. doi:10.1016/bs.dnb.2024.02.005 , abstract =

    work page doi:10.1016/bs.dnb.2024.02.005 2024

  30. [32]

    AI and Ethics , author =

    The ethics of. AI and Ethics , author =. 2025 , keywords =. doi:10.1007/s43681-025-00815-4 , abstract =

    work page doi:10.1007/s43681-025-00815-4 2025

  31. [33]

    JMIR mHealth and uHealth , author =

    Ethical. JMIR mHealth and uHealth , author =. 2021 , pages =. doi:10.2196/27343 , abstract =

    work page doi:10.2196/27343 2021

  32. [34]

    preprint , author =

    The. preprint , author =. doi:10.17605/OSF.IO/5AKEB , abstract =

    work page doi:10.17605/osf.io/5akeb

  33. [35]

    Bufano, Pasquale and Laurino, Marco and Said, Sara and Tognetti, Alessandro and Menicucci, Danilo , month = dec, year =. Digital. Journal of Medical Internet Research , publisher =. doi:10.2196/46778 , abstract =

    doi:10.2196/46778

  34. [36]

    preprint , author =

    The. preprint , author =

  35. [37]

    Yahoo Finance , author =

    42\. Yahoo Finance , author =

  36. [38]

    Psychiatric Services , author =

    Financial. Psychiatric Services , author =. 2013 , pages =. doi:10.1176/appi.ps.005882012 , language =

    work page doi:10.1176/appi.ps.005882012 2013

  37. [39]

    arXiv.org , author =

    Using. arXiv.org , author =

  38. [40]

    American Journal of Physical Medicine & Rehabilitation , author =

    Digital. American Journal of Physical Medicine & Rehabilitation , author =. 2020 , pages =. doi:10.1097/PHM.0000000000001506 , abstract =

    work page doi:10.1097/phm.0000000000001506 2020

  39. [41]

    and Camelo, Evelyn V

    Silva, Rafael de Assis da and Mograbi, Daniel C. and Camelo, Evelyn V. M. and Bifano, Jaqueline and Wainstok, Mayra and Silveira, Luciana Angélica Silva and Cheniaux, Elie , year =. Insight in bipolar disorder: a comparison between mania, depression and euthymia using the. Trends in Psychiatry and Psychotherapy , publisher =. doi:https://doi.org/10.1590/2...

    work page doi:10.1590/2237-6089-2015-0014 2015

  40. [42]

    The Psychiatric quarterly , author =

    Insight in. The Psychiatric quarterly , author =. 2011 , pages =. doi:10.1007/s11126-011-9200-4 , abstract =

    work page doi:10.1007/s11126-011-9200-4 2011

  41. [43]

    and Gruber, June , month = sep, year =

    Giovanelli, Alison and Hoerger, Michael and Johnson, Sheri L. and Gruber, June , month = sep, year =. Impulsive responses to positive mood and reward are related to mania risk , volume =. Cognition and Emotion , publisher =. doi:10.1080/02699931.2013.772048 , abstract =

    arXiv 2013

  42. [44]

    JMIR Mental Health , author =

    Clinician. JMIR Mental Health , author =. 2021 , pages =. doi:10.2196/25455 , abstract =

    work page doi:10.2196/25455 2021

  43. [45]

    Wang, Xiaomei and Smith, Alec and Keller, Bruce and Sasangohar, Farzan , month = jun, year =. Mobile. doi:10.48550/arXiv.2206.02960 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2206.02960

  44. [46]

    and Abrami, Avner and Heisig, Stephen and Ali, Asra and Arenare, Elizabeth and Agurto, Carla and Lu, Nathaniel and Kane, John M

    Birnbaum, Michael L. and Abrami, Avner and Heisig, Stephen and Ali, Asra and Arenare, Elizabeth and Agurto, Carla and Lu, Nathaniel and Kane, John M. and Cecchi, Guillermo , month = jan, year =. Acoustic and. JMIR Mental Health , publisher =. doi:10.2196/24699 , abstract =

    work page doi:10.2196/24699

  45. [47]

    , month = jun, year =

    Onnela, Jukka-Pekka and Rauch, Scott L. , month = jun, year =. Harnessing. Neuropsychopharmacology , publisher =. doi:10.1038/npp.2016.7 , language =

    doi:10.1038/npp.2016.7 2016

  46. [48]

    Chandrasekara, Gayan and G.K., Munasinghe and K.K.U.M., Hansani and M.T.L., Jayawardhana , month = dec, year =. User-. doi:10.13140/RG.2.2.11024.11529 , abstract =

    arXiv

  47. [49]

    Human-crafted

    Liang Tang and Nigel Bosch , editor =. Human-crafted. doi:10.5281/ZENODO.15870199 , abstract =

    work page doi:10.5281/zenodo.15870199

  48. [50]

    , volume =

    Implementation strategies for digital mental health interventions in health care settings. , volume =. American Psychologist , author =. 2020 , pages =. doi:10.1037/amp0000686 , abstract =

    work page doi:10.1037/amp0000686 2020

  49. [51]

    Thematic analysis , volume =

    Clarke, Victoria and Braun, Virginia , month = may, year =. Thematic analysis , volume =. The Journal of Positive Psychology , publisher =. doi:10.1080/17439760.2016.1262613 , number =

    arXiv 2016

  50. [52]

    Individual Behavioral Insights in Schizophrenia: A Network Analysis and Mobile Sensing Approach

    Davies, Andy and Fried, Eiko and Aung, Hane and Costilla-Reyes, Omar , month = dec, year =. Individual. doi:10.48550/arXiv.2312.01216 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2312.01216

  51. [53]

    One size fits all?

    Braun, Virginia and Clarke, Victoria , month = jul, year =. One size fits all?. Qualitative Research in Psychology , publisher =. doi:10.1080/14780887.2020.1769238 , abstract =

    arXiv 2020

  52. [54]

    Health Economics , author =

    Stated versus revealed preferences:. Health Economics , author =. 2021 , keywords =. doi:10.1002/hec.4246 , abstract =

    doi:10.1002/hec.4246 2021

  53. [55]

    arXiv.org , author =

    Human-. arXiv.org , author =

  54. [56]

    JAMA Internal Medicine , author =

    Financial. JAMA Internal Medicine , author =. 2021 , pages =. doi:10.1001/jamainternmed.2020.6432 , abstract =

    arXiv 2021

  55. [57]

    A systematic review of early warning systems in finance

    Namaki, Ali and Eyvazloo, Reza and Ramtinnia, Shahin , month = sep, year =. A systematic review of early warning systems in finance , url =. doi:10.48550/arXiv.2310.00490 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2310.00490

  56. [58]

    Economic Modelling , author =

    Early warning models against bankruptcy risk for. Economic Modelling , author =. 2013 , keywords =. doi:10.1016/j.econmod.2012.11.017 , abstract =

    work page doi:10.1016/j.econmod.2012.11.017 2013

  57. [59]

    National Institute on Aging , month = feb, year =

    Difficulty managing bills may signal early dementia , url =. National Institute on Aging , month = feb, year =

  58. [60]

    JAMA Network Open , author =

    Early. JAMA Network Open , author =. 2025 , pages =. doi:10.1001/jamanetworkopen.2025.15894 , abstract =

    arXiv 2025

  59. [61]

    Cogent Social Sciences , author =

    Reshaping regulatory governance: leveraging. Cogent Social Sciences , author =. 2025 , pages =. doi:10.1080/23311886.2025.2474193 , abstract =

    arXiv 2025

  60. [62]

    Fintech for

    Brozena, Jeff and Blair, Johnna and Richardson, Thomas and Matthews, Mark and Mukherjee, Dahlia and Saunders, Erika F H and Abdullah, Saeed , year =. Fintech for

  61. [63]

    Dlima, Schenelle Dayna and Shevade, Santosh and Menezes, Sonia Rebecca and Ganju, Aakash , month = jul, year =. Digital. JMIR Bioinformatics and Biotechnology , publisher =. doi:10.2196/39618 , abstract =

    work page doi:10.2196/39618

  62. [64]

    and Zin, May The and Boonstra, Tjeerd W

    Braund, Taylor A. and Zin, May The and Boonstra, Tjeerd W. and Wong, Quincy J. J. and Larsen, Mark E. and Christensen, Helen and Tillman, Gabriel and O’Dea, Bridianne , month = may, year =. Smartphone. JMIR Mental Health , publisher =. doi:10.2196/35549 , abstract =

    work page doi:10.2196/35549

  63. [65]

    Beiwinkel, Till and Kindermann, Sally and Maier, Andreas and Kerl, Christopher and Moock, Jörn and Barbian, Guido and Rössler, Wulf , month = jan, year =. Using. JMIR Mental Health , publisher =. doi:10.2196/mental.4560 , abstract =

    work page doi:10.2196/mental.4560

  64. [66]

    Archives of General Psychiatry , author =

    Lifetime and 12-. Archives of General Psychiatry , author =. 2007 , pages =. doi:10.1001/archpsyc.64.5.543 , abstract =

    work page doi:10.1001/archpsyc.64.5.543 2007

  65. [67]

    AARP , author =

    Older. AARP , author =. doi:10.26419/res.00928.001 , abstract =

    work page doi:10.26419/res.00928.001

  66. [68]

    The role of financial

  67. [69]

    The future of fintech growth , url =

  68. [70]

    Journal of Psychiatric Research , author =

    The prevalence and features of schizophrenia among individuals with gambling disorder , volume =. Journal of Psychiatric Research , author =. 2021 , keywords =. doi:10.1016/j.jpsychires.2021.02.025 , abstract =

    work page doi:10.1016/j.jpsychires.2021.02.025 2021

  69. [72]

    Gupta, Ankit and Heng, Tim and Shaw, Chris and Gromala, Diane and Leese, Jenny and Li, Linda , month = feb, year =. Oh,. Proceedings of the 14th. doi:10.1145/3421937.3421991 , abstract =

    arXiv

  70. [73]

    doi:10.12688/f1000research.157584.1 , abstract =

    Ivezić, Dijana and Keppel, Jonas and Horneber, David and Becker, Christine and Laumer, Sven and Walle, Hardy and Schneegass, Stefan and Amft, Oliver , month = nov, year =. doi:10.12688/f1000research.157584.1 , abstract =

    work page doi:10.12688/f1000research.157584.1

  71. [74]

    Advances in Health Sciences Education , author =

    Likert scales, levels of measurement and the “laws” of statistics , volume =. Advances in Health Sciences Education , author =. 2010 , keywords =. doi:10.1007/s10459-010-9222-y , abstract =

    doi:10.1007/s10459-010-9222-y 2010

  72. [75]

    Associations between wearables vital parameters and self-perceived mood—an ecological momentary assessment study among healthy adolescents , volume =

    Lutz, Luisa and Veltmann, Polina and Meisterhans, Romain and Giurgiu, Marco , month = dec, year =. Associations between wearables vital parameters and self-perceived mood—an ecological momentary assessment study among healthy adolescents , volume =. Frontiers in Psychology , publisher =. doi:10.3389/fpsyg.2025.1623886 , abstract =

    arXiv 2025

  73. [77]

    International Journal of Psychophysiology , author =

    Heart rate variability as a potential indicator of positive valence system disturbance:. International Journal of Psychophysiology , author =. 2015 , keywords =. doi:10.1016/j.ijpsycho.2015.08.005 , abstract =

    work page doi:10.1016/j.ijpsycho.2015.08.005 2015

  74. [78]

    Sensors , author =

    Heart. Sensors , author =. 2023 , keywords =. doi:10.3390/s23020966 , abstract =

    work page doi:10.3390/s23020966 2023

  75. [79]

    Associations

    Hannon, James and O’Hagan, Adrian and Lambe, Rory and O’Grady, Ben and Doherty, Cailbhe , month = jan, year =. Associations. Sensors , publisher =. doi:10.3390/s25144415 , abstract =

    work page doi:10.3390/s25144415

  76. [80]

    JMIR Formative Research , author =

    Assessment of. JMIR Formative Research , author =. 2024 , pages =. doi:10.2196/49396 , abstract =

    work page doi:10.2196/49396 2024

  77. [81]

    , month = mar, year =

    Yang, Jessica and Kershaw, Kiarri N. , month = mar, year =. Feasibility of using ecological momentary assessment and continuous heart rate monitoring to measure stress reactivity in natural settings , volume =. PLOS ONE , publisher =. doi:10.1371/journal.pone.0264200 , abstract =

    work page doi:10.1371/journal.pone.0264200

  78. [82]

    Evaluation Review , author =

    Centering or. Evaluation Review , author =. 2006 , pages =. doi:10.1177/0193841X05275649 , abstract =

    work page doi:10.1177/0193841x05275649 2006

  79. [83]

    Current Cardiology Reports , author =

    Associations. Current Cardiology Reports , author =. 2025 , keywords =. doi:10.1007/s11886-025-02299-4 , abstract =

    work page doi:10.1007/s11886-025-02299-4 2025

  80. [84]

    International Journal of Bipolar Disorders , author =

    The futility of long-term predictions in bipolar disorder: mood fluctuations are the result of deterministic chaotic processes , volume =. International Journal of Bipolar Disorders , author =. 2021 , keywords =. doi:10.1186/s40345-021-00235-3 , abstract =

    work page doi:10.1186/s40345-021-00235-3 2021

Showing first 80 references.