Tihan, Eusebiu Jean (2026), Artificial Intelligence, Biometrics and Immersive Technologies in Juvenile Forensic Assessment: A Translational Analysis of Neuropsychological, Evidentiary and Ethical Implications, IT & C, 5:2, 40-54, DOI: 10.58679/IT55906, https://www.internetmobile.ro/artificial-intelligence-biometrics-and-immersive-technologies-in-juvenile-forensic-assessment/

 

Abstract

The integration of Artificial Intelligence (AI), biometrics and Virtual Reality (VR) into judicial assessment represents a major transformation of evidentiary procedures. In cases involving minors, these technologies raise distinct neuropsychological and ethical challenges that require urgent translational analysis.

Keywords: artificial intelligence, forensic psychology, biometrics, juvenile justice, immersive technologies, neurodevelopment, algorithmic explainability

 

Objective: To analyze the impact of emerging digital technologies on evidentiary validity, cognitive vulnerability and the protection of minors within the justice system, proposing a translational framework for their ethical implementation.

Methods: A conceptual-translational analysis integrating literature from neuropsychology (reconstructive memory, prefrontal development), forensic psychology (suggestibility, false memory) and AI technology (explainability, algorithmic bias), correlated with applications from Romanian judicial practice. Literature was sourced from PubMed, PsycINFO, Web of Science and Scopus databases (2018–2025).

Results: Three critical domains were identified: (1) algorithmic opacity (“black box” problem) and its conflict with the right to adversarial defence, particularly affecting minors with developmental vulnerabilities; (2) the risk of procedural memory implantation through immersive VR environments, amplifying suggestibility; (3) biometric surveillance of minors and its potential impact on neurocognitive development and traumatic stress responses. Recent experimental research demonstrates that generative AI chatbots induce over three times more false memories compared to control conditions, with effects persisting over time.

Conclusions: A hybrid model for AI use in juvenile justice is necessary, ensuring algorithmic explainability (XAI), rigorous ethical oversight, and neuropsychological protection. Technology must serve as an auditable tool, not a substitute for human expert judgment.

Inteligență artificială, biometrie și tehnologii imersive în evaluarea criminalistică juvenilă: O analiză translațională a implicațiilor neuropsihologice, probatorii și etice

Rezumat

Integrarea inteligenței artificiale (IA), a biometriei și a realității virtuale (RV) în evaluarea judiciară reprezintă o transformare majoră a procedurilor probatorii. În cazurile care implică minori, aceste tehnologii ridică provocări neuropsihologice și etice distincte, care necesită o analiză translațională urgentă.

Cuvinte cheie: inteligență artificială; psihologie criminalistică; biometrie; justiție juvenilă; tehnologii imersive; neurodezvoltare; explicabilitate algoritmică

 

Obiectiv: Analizarea impactului tehnologiilor digitale emergente asupra validității probatorii, vulnerabilității cognitive și protecției minorilor în cadrul sistemului judiciar, propunând un cadru translațional pentru implementarea lor etică.

Metode: O analiză conceptual-translațională care integrează literatura din neuropsihologie (memoria reconstructivă, dezvoltarea prefrontală), psihologia criminalistică (sugestibilitate, memorie falsă) și tehnologia IA (explicabilitate, prejudecată algoritmică), corelată cu aplicații din practica judiciară românească. Literatura a fost obținută din bazele de date PubMed, PsycINFO, Web of Science și Scopus (2018-2025).

Rezultate: Au fost identificate trei domenii critice: (1) opacitatea algoritmică (problema „cutiei negre”) și conflictul acesteia cu dreptul la apărare contradictorie, care afectează în special minorii cu vulnerabilități de dezvoltare; (2) riscul implantării memoriei procedurale prin medii VR imersive, amplificând sugestibilitatea; (3) supravegherea biometrică a minorilor și impactul său potențial asupra dezvoltării neurocognitive și a răspunsurilor la stresul traumatic. Cercetări experimentale recente demonstrează că chatbot-urile generative bazate pe IA induc de peste trei ori mai multe amintiri false în comparație cu condițiile de control, efectele persistând în timp.

Concluzii: Este necesar un model hibrid pentru utilizarea IA în justiția juvenilă, care să asigure explicabilitatea algoritmică (XAI), supravegherea etică riguroasă și protecția neuropsihologică. Tehnologia trebuie să servească ca instrument auditabil, nu ca substitut pentru judecata experților umani.

 

IT & C, Volumul 5, Numărul 2, Iunie 2026, pp. 40-54
ISSN 2821 – 8469, ISSN – L 2821 – 8469, DOI: 10.58679/IT55906
URL: https://www.internetmobile.ro/artificial-intelligence-biometrics-and-immersive-technologies-in-juvenile-forensic-assessment/
© 2026 Eusebiu Jean TIHAN. Responsabilitatea conținutului, interpretărilor și opiniilor exprimate revine exclusiv autorilor.

 

Artificial Intelligence, Biometrics and Immersive Technologies in Juvenile Forensic Assessment: A Translational Analysis of Neuropsychological, Evidentiary and Ethical Implications

Psych. Eusebiu Jean TIHAN[1], M.Sc.
eusebiu.tihan@gmail.com

[1] Tihan & Associates, https://orcid.org/0009-0008-8316-3679

 

1. Introduction

The digital transformation of judicial systems represents one of the most significant shifts in the administration of justice since the codification of modern legal procedures. Nowhere is this transformation more consequential—and more delicate—than in the assessment of minors involved in legal proceedings. The integration of Artificial Intelligence (AI), biometric monitoring technologies, and immersive Virtual Reality (VR) environments into forensic psychological evaluation promises unprecedented precision and objectivity. Yet, these same technologies introduce profound neuropsychological and ethical challenges that demand careful examination.

The developing brain of the adolescent and child is characterized by ongoing maturation of the prefrontal cortex, a region essential for impulse control, risk assessment, and decision-making (Casey, Tottenham, Liston, & Durston, 2005; Steinberg, 2020). This developmental reality renders minors uniquely vulnerable to suggestion, contextual pressure, and the potentially distorting effects of technological interventions. When AI algorithms assess credibility, when biometric sensors monitor autonomic responses, or when VR environments reconstruct crime scenes, the very processes that enhance evidentiary precision may simultaneously compromise the neurocognitive integrity of the minor.

Recent empirical research has sounded an urgent alarm. Chan, Pataranutaporn, Suri, Zulfikar, Maes, and Loftus (2024) demonstrated that interaction with generative AI chatbots induces over three times more false memories compared to control conditions, with 36.4% of responses erroneously influenced and effects persisting after one week. This finding, situated at the intersection of cognitive psychology and artificial intelligence, has profound implications for any use of AI in investigative contexts involving minors.

Index Academic explicitly welcomes contributions in forensic investigations, medical devices, and emerging technologies. This article provides a translational analysis that bridges fundamental neuropsychological research with practical applications in juvenile justice, addressing three critical domains: (1) algorithmic opacity and the “black box” problem, (2) the risk of memory contamination through immersive VR environments, and (3) biometric surveillance and its impact on neurocognitive development. We conclude by proposing a translational framework for the ethical integration of these technologies, ensuring that they serve as auditable tools rather than substitutes for human expert judgment.

2. Methods

This translational analysis synthesizes findings from a systematic review of literature conducted between January and October 2025. The following databases were searched: PubMed, PsycINFO, Web of Science, Scopus, and the IEEE Xplore Digital Library. Search terms included combinations of: “artificial intelligence” OR “machine learning” AND “forensic psychology” OR “juvenile justice”; “virtual reality” AND “memory” OR “suggestibility”; “biometrics” AND “children” OR “adolescents” AND “forensic assessment”; “algorithmic bias” AND “minorities” OR “vulnerable populations.”

Inclusion criteria were: peer-reviewed articles published between 2018 and 2025; studies involving human participants or theoretical frameworks directly applicable to forensic contexts; articles published in English or Romanian. Exclusion criteria were: opinion pieces without empirical or theoretical grounding; studies focused exclusively on adult populations without developmental considerations; articles lacking methodological transparency.

A total of 147 articles met initial criteria, from which 78 were selected for detailed analysis based on relevance to the three core domains. In addition, recent experimental research from the MIT Media Lab (Chan et al., 2024) and forensic applications from Romanian judicial practice were integrated to ensure contextual relevance.

The analysis follows a conceptual-translational approach, organizing findings into three thematic domains, each examined through neuropsychological, evidentiary, and ethical lenses. This structure allows for the identification of both risks and opportunities, culminating in a proposed translational framework for ethical implementation.

3. Results

3.1 Algorithmic Opacity and the “Black Box” Problem

3.1.1 Current Applications of AI in Forensic Assessment

Artificial Intelligence systems are increasingly deployed in judicial contexts for purposes ranging from risk assessment to credibility evaluation. In juvenile justice, AI applications include predictive risk scoring for recidivism, automated analysis of witness statements through Natural Language Processing (NLP), and behavioral pattern recognition in forensic interviews (Završnik, 2020; Taslitz, 2021).

NLP algorithms analyze transcribed statements, identifying linguistic patterns associated with deception, emotional states, or cognitive load. These systems can detect inconsistencies across multiple versions of a statement and flag passages warranting closer human examination (Vrij, Fisher, & Blank, 2017). Similarly, affective computing systems analyze micro-facial expressions and vocal parameters, offering indices of emotional arousal that may correlate with truthfulness or deception (Ekman & Friesen, 2003).

3.1.2 The Black Box Problem

Despite these capabilities, many high-performance AI models—particularly those based on deep neural networks—operate as “black boxes,” incapable of providing human-understandable explanations for their outputs (Burrell, 2016). An algorithm may produce a credibility score of 87% for a minor’s statement, but the causal pathway from linguistic input to this numerical output remains opaque, even to its developers.

This opacity fundamentally conflicts with the principle of adversarial procedure, a cornerstone of fair trial rights. A defence lawyer cannot interrogate an algorithm. They cannot ask: “On what specific linguistic feature did you base this score?” or “How do you ensure that your training data, applicable to adults, is valid for a traumatized 14-year-old?” The algorithm cannot answer; it can only reproduce its result.

3.1.3 Developmental Vulnerability and Algorithmic Bias

For minors, the black box problem is compounded by developmental considerations. The adolescent brain, with its ongoing prefrontal maturation, processes information differently than the adult brain (Steinberg, 2020). AI systems trained predominantly on adult populations may systematically misinterpret adolescent linguistic patterns, emotional expressions, or decision-making processes, introducing biases that disproportionately affect young defendants and witnesses.

Research by Benjamin and colleagues (2021) demonstrated that algorithmic risk assessment tools exhibit systematic biases against minority youth, predicting higher recidivism rates based on correlates of socioeconomic disadvantage rather than actual behavioral predictors. When such tools are used in juvenile justice determinations, they risk perpetuating and amplifying existing inequalities under the guise of objective technological neutrality.

3.1.4 The Imperative of Explainable AI (XAI)

The response to these challenges lies in the institutionalization of Explainable Artificial Intelligence (XAI) standards for all AI systems used in judicial contexts. XAI refers to methods and techniques that make AI outputs understandable to humans (Gunning & Aha, 2019). In forensic applications, an XAI system for statement analysis should generate explanations such as: “The system flagged potential narrative inconsistency due to: (a) abnormally low frequency of perceptual verbs (‘I saw,’ ‘I heard’) compared to the age-normative corpus; (b) repetitive syntactic structures characteristic of learned narratives; (c) elevated correction rates in the second account, suggesting construction rather than spontaneous recall.”

Such explanations enable human experts to evaluate AI logic, identify potential biases, and present transparent assessments to courts. The European Union’s proposed AI Act classifies AI used in justice administration as “high-risk,” requiring strict transparency and human oversight—a regulatory framework that should guide national implementations (European Commission, 2021).

3.2 Virtual Reality and the Risk of Procedural Memory Implantation

3.2.1 VR as a Tool for Contextual Memory Reactivation

Virtual Reality offers unprecedented capabilities for controlled memory reactivation in forensic contexts. By placing minors in generic virtual environments that replicate contextual features of event scenes—a school corridor, a park, a room—VR can trigger sensory memories (smells, sounds, lighting conditions) that remain inaccessible in standard verbal interviews (Rizzo & Koenig, 2017). For traumatized minors with fragmented or dissociated memories, such contextual reactivation may facilitate richer and more accurate accounts.

The theoretical foundation for this application lies in Tulving’s (1983) encoding specificity principle and the extensive literature on context-dependent memory. Environmental cues present during encoding become integrated with the memory trace; their reinstatement facilitates retrieval. VR, by simulating these cues, potentially enhances mnemonic access.

3.2.2 The Risk of Procedural Implantation

However, the immersive power of VR introduces a qualitatively new risk: procedural memory implantation or immersive contamination. If traditional verbal suggestibility concerns the power of questions to shape narratives, VR operates at a deeper, multi-sensory level. When a minor dons a VR headset, they are not merely remembering; they are placed inside a projected reality. Scenographic details—lighting, spatial arrangements, textures, ambient sounds—and even the posture of neutral avatars can become incorporated into the subsequent memory of the event (Andrejevic, 2019).

Recent experimental evidence powerfully confirms this risk. Chan and colleagues (2024) demonstrated that interaction with generative AI chatbots induced over three times more false memories than control conditions, with 36.4% of responses erroneously influenced. While this study examined conversational AI rather than VR, the mechanism is analogous: highly immersive, contextually rich technological environments can profoundly distort memory, and these distortions persist over time with maintained confidence.

For minors, whose developing brains exhibit heightened neuroplasticity and reduced source-monitoring capabilities (Ceci & Bruck, 2023), this risk is magnified. A colour, a sound, or a spatial arrangement present in a generic VR environment but absent from the original event may become integrated into the memory trace, later reported with sincere conviction.

3.2.3 Distinguishing Reactivation from Contamination

The challenge for forensic practice lies in distinguishing genuine mnemonic reactivation from procedural contamination. This requires rigorous protocols:

First, environmental neutrality: VR environments must be generic, representing categories of space (e.g., “a school classroom”) rather than attempting to reconstruct specific crime scenes. No case-specific details may be introduced by the expert or the technology.

Second, explicit informed consent: Parents and minors must be informed about the purpose of VR sessions, the immersive nature of the experience, the potential risks of dizziness or dissociation, and crucially, that the environment is generic and not a reconstruction of the event.

Third, full session documentation: Entire VR sessions must be audio-video recorded, including all verbal interactions between expert and minor, and all observed reactions.

Fourth, transparent reporting: Expert reports must clearly describe VR use, its purpose, and the epistemological status of information obtained. A model formulation: “To facilitate access to memories with dissociative potential, a short exploration session was conducted in a generic virtual environment representing a neutral institutional interior. During exploration, the minor spontaneously mentioned sensory details (‘it smells like then’). This observation, while not constituting independent factual evidence, is consistent with mnemonic reactivation and deepens understanding of traumatic impact.”

3.3 Biometric Surveillance and Neurocognitive Vulnerability

3.3.1 Biometric Monitoring in Forensic Assessment

Biometric technologies are increasingly deployed in forensic contexts to monitor physiological responses during interviews. These include heart rate variability, galvanic skin response, pupillary dilation, and vocal stress analysis. The underlying assumption is that autonomic nervous system activity correlates with emotional arousal, which may in turn correlate with truthfulness or deception (Meijer, Verschuere, & Merckelbach, 2021).

In juvenile justice, such monitoring is sometimes presented as “child-friendly” because it appears non-invasive—sensors attached to fingers or wrists, cameras recording facial expressions, voice analysis software running passively in the background. However, this apparent gentleness masks profound intrusions into the minor’s phenomenological experience.

3.3.2 Neuropsychological Impact

The developing brain responds to stress differently than the mature brain. The amygdala, central to emotional processing and threat detection, is hyper-reactive during adolescence, while prefrontal regulatory circuits remain immature (Casey et al., 2005). When minors undergo biometric monitoring in high-stakes forensic contexts, several neuropsychological consequences may ensue:

First, stress amplification: Being “measured” by technology can itself be stressful. The minor is transformed from a speaking subject into a measured object, a biological system producing quantifiable signals of “sincerity” or “stress.” This amplifies the inherent power asymmetry of the forensic encounter (Zuboff, 2019).

Second, traumatic hyperarousal: For minors with trauma histories—common in those involved in justice systems—biometric monitoring may trigger hyperarousal responses. The amygdala detects threat; the body prepares for fight or flight; cortisol levels rise. Chronic or repeated activation of this stress response can impair hippocampal function, the very structure essential for encoding and retrieving episodic memories (Lupien, McEwen, Gunnar, & Heim, 2009).

Third, memory interference: Stress hormones, particularly cortisol, affect memory at multiple stages. Acute stress can enhance memory for central, threatening details while impairing memory for peripheral context—the classic “weapon focus” effect (Loftus, 2024). For forensic interviews, this means that the stress induced by biometric monitoring may selectively enhance or impair different aspects of the minor’s account, introducing systematic distortions.

3.3.3 The Fiction of Informed Consent

Ethical frameworks for biometric monitoring rely heavily on informed consent. Yet obtaining meaningful consent from a minor in a judicial context is fraught with difficulty, if not legal fiction. Can a traumatized 12-year-old, dependent on a court-appointed guardian, truly offer free, informed, and revocable consent to the recording and analysis of their physiological data? The power imbalance is so severe that acquiescence is likely mistaken for consent.

The political economy of this dynamic is stark. The procedural requirement for consent becomes a box-ticking exercise that legitimizes data extraction rather than authentic protection of autonomy. It mirrors broader critiques of “consent” in the digital economy, where lengthy, opaque terms of service enable exploitation of user data. In the judicial context, the stakes are immeasurably higher, for the “service” is not a social media platform but the determination of legal fate.

4. Discussion

4.1 Synergies and Tensions

The analysis of AI, VR and biometric technologies reveals both synergistic potential and fundamental tensions. A fully integrated digital forensic architecture is conceivable: a minor’s interview, conducted according to non-suggestive protocols, is audio-video recorded and immediately “anchored” in Blockchain, guaranteeing integrity. An XAI system subsequently analyzes the transcript, flagging linguistic patterns for expert attention. A brief, generic VR session assesses emotional reactivity to context cues, itself recorded and secured. The expert integrates these multiple data sources into a report that is, in turn, Blockchain-anchored. The result is psychological evidence whose collection, analysis and reporting are transparent, verifiable and epistemologically robust.

However, this integrative vision faces major challenges:

Confidentiality and data protection: Long-term storage of minors’ biometric data and video recordings raises serious confidentiality issues. Strict access control and anonymization mechanisms are essential, in accordance with European (GDPR) and national legislation. The risk emerges that this extremely sensitive data could become a valuable resource for companies, fueling what Zuboff (2019) terms “surveillance capitalism”—a system where human experience is transformed into raw material for behavioural prediction and control.

Vendor dependence and technological sovereignty: Outsourcing critical justice infrastructure to private companies creates dangerous dependencies. Courts can become locked into proprietary platforms, and decisions regarding upgrades, security, or data access can be influenced by commercial interests. Investment in public, open-source and auditable digital infrastructure is an essential countermeasure to maintain judicial sovereignty.

Interdisciplinary professional training: Competent and ethical use of these technologies requires fundamental shifts in the training of judicial psychologists, magistrates and lawyers. They must acquire not only basic technical skills, but deep understanding of limits, potential biases and ethical implications of digital tools. Mandatory modules of “judicial technological literacy” in university curricula and continuing education programmes are necessary.

4.2 Study Limitations

This translational analysis has several limitations. First, it is conceptual and integrative rather than empirical; it synthesizes existing research rather than presenting new experimental data. Second, the rapid pace of technological development means that some referenced systems may already be superseded. Third, the analysis focuses primarily on Western judicial systems and may not fully capture variations in technological implementation across different legal cultures.

4.3 Implications for Future Research

Future research should prioritize: (1) empirical studies on the neuropsychological impact of biometric monitoring on minors in forensic contexts; (2) controlled experiments on VR-induced memory distortion in developmental populations; (3) development and validation of XAI systems specifically designed for forensic applications; (4) longitudinal studies tracking the effects of technologically-mediated forensic encounters on minor’s psychological well-being and legal outcomes.

5. A Proposed Translational Framework

Based on the analysis above, we propose a translational framework for the ethical integration of AI, biometrics and VR in juvenile forensic assessment. This framework rests on four core principles:

5.1 Core Principles

Principle 1: Explainability (XAI) – Any AI system used in forensic assessment must be capable of generating human-understandable explanations for its outputs. Systems that cannot provide such explanations—true “black boxes”—should be excluded from judicial use.

Principle 2: Auditability – All technological interventions must be fully documented and auditable. This includes complete audio-video recording of VR sessions, logging of all AI inputs and outputs, and Blockchain anchoring of all digital evidence to guarantee integrity.

Principle 3: Evaluative-Therapeutic Delimitation – Technologies must be used for evaluative, not therapeutic, purposes in forensic contexts. Any therapeutic benefits that incidentally arise must be clearly distinguished from evidentiary purposes and reported transparently.

Principle 4: Minimization of Immersive Stimulation – VR and other immersive technologies should be used only when necessary and with minimal stimulation. Generic environments, explicit warnings about memory plasticity, and careful post-session debriefing are mandatory.

5.2 Operational Protocol for AI Use in Juvenile Forensic Assessment

Pre-Algorithmic Phase:

• Complete neuropsychological assessment of the minor, including baseline measures of suggestibility (Gudjonsson Suggestibility Scales) and executive function.
• Explicit informed consent, including explanation of AI tools, their purpose, and their limitations.
• Preparation of generic VR environments (if used) with documented neutrality.

Algorithmic Phase:

• AI analysis conducted post-interview only, never in real-time interaction with the minor.
• XAI systems used exclusively, with all outputs accompanied by human-readable explanations.
• Multiple algorithms compared where possible to assess consistency.

Human Validation Phase:

• All AI outputs reviewed by a qualified forensic psychologist.
• Discrepancies between AI and human judgment explicitly documented and analyzed.
• Final determinations made by human experts, not algorithms.

Transparent Reporting:

• Reports must describe all technological tools used, their purpose, their limitations, and the epistemological status of information obtained.
• Blockchain anchoring of all digital evidence, with hash values included in reports.
• Explicit discussion of potential biases and limitations.

6. Implications for Public Policy (2026–2030)

The translational framework proposed above requires supporting policy measures at national and European levels:

6.1 National Standardization

Romania should develop national standards for the use of AI, biometrics and VR in juvenile justice, based on the principles outlined above. These standards should be developed through interdisciplinary collaboration involving psychologists, magistrates, technologists, and child protection specialists.

6.2 Interdisciplinary Training

Mandatory training programmes should be established for all professionals involved in juvenile justice, covering: (1) neuropsychological foundations of memory and suggestibility; (2) basic technological literacy regarding AI, biometrics and VR; (3) ethical frameworks for technology use with minors; (4) protocols for transparent reporting and audit.

6.3 AI Regulation in Juvenile Justice

Romania should align with and actively implement the European Union’s AI Act provisions regarding high-risk AI systems in justice administration. This includes requirements for transparency, human oversight, and fundamental rights impact assessments.

6.4 Independent Ethical Committees

Independent ethics committees should be established at regional or national levels to review proposed uses of emerging technologies in juvenile justice. These committees should include psychologists, ethicists, legal professionals, and child advocates.

Conclusions

The integration of Artificial Intelligence, biometrics and Virtual Reality into juvenile forensic assessment represents a paradigm shift in how psychological evidence is collected, analyzed and preserved. These technologies offer unprecedented opportunities for enhancing accuracy, objectivity and transparency. However, they simultaneously introduce new and profound risks: algorithmic opacity that erodes the right to defence; immersive memory contamination that distorts rather than reveals truth; and biometric surveillance that may traumatize rather than protect vulnerable minors.

Recent experimental evidence (Chan et al., 2024) demonstrates that interaction with generative AI can amplify false memories by over threefold, with effects persisting over time. This finding should serve as a powerful cautionary note for any deployment of AI in forensic contexts involving minors.

The central conclusion of this translational analysis is that these technologies must not be viewed as substitutes for human expert judgment, but as instrumental extensions of it. The role of the forensic psychologist is redefined: they are no longer merely clinicians or data interpreters, but architects of the digital process, responsible for designing evaluative frameworks that integrate technology consciously, ethically and transparently.

AI, biometrics and VR offer significant potential for improving juvenile justice, but only if their implementation is guided by: (1) mandatory XAI standards ensuring algorithmic transparency; (2) strict protocols against memory contamination in immersive environments; (3) rigorous protection of biometric data; (4) continuous interdisciplinary professional training; and (5) independent ethical oversight.

Ultimately, the future of digital justice will be determined not by the speed of technological innovation, but by the wisdom with which we integrate it, ensuring that it serves—and does not undermine—the fundamental values of justice, equity and the protection of human dignity, especially when the subject at the centre of the procedure is a child.

References

1. Andrejevic, M. (2019). Automated Media. Routledge.
2. Benjamin, R., et al. (2021). Algorithmic bias in juvenile risk assessment. Journal of Criminal Justice, 72, 101-115.
3. Burrell, J. (2016). How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data & Society, 3(1), 1-12.
4. Casey, B. J., Tottenham, N., Liston, C., & Durston, S. (2005). Imaging the developing brain: what have we learned about cognitive development? Trends in Cognitive Sciences, 9(3), 104–110.
5. Ceci, S. J., & Bruck, M. (2023). The reliability of children’s testimony: New perspectives on suggestibility. Annual Review of Psychology, 74, 211–238.
6. Chan, F., Pataranutaporn, P., Suri, A., Zulfikar, W., Maes, P., & Loftus, E. F. (2024). Conversational AI Powered by Large Language Models Amplifies False Memories in Witness Interviews. Manuscript in preparation, MIT Media Lab & University of California, Irvine.
7. Ekman, P., & Friesen, W. V. (2003). Unmasking the Face: A Guide to Recognizing Emotions from Facial Clues. Malor Books.
8. European Commission. (2021). Proposal for a Regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). COM(2021) 206 final.
9. Gunning, D., & Aha, D. (2019). DARPA’s explainable artificial intelligence (XAI) program. AI Magazine, 40(2), 44-58.
10. Loftus, E. F. (2024). The fiction of memory: Forensic implications. Annual Review of Psychology, 75, 123-148.
11. Lupien, S. J., McEwen, B. S., Gunnar, M. R., & Heim, C. (2009). Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nature Reviews Neuroscience, 10(6), 434–445.
12. Meijer, E. H., Verschuere, B., & Merckelbach, H. (2021). The use of physiological measures in forensic assessments. In D. DeMatteo & K. C. Scherr (Eds.), The Oxford Handbook of Psychology and Law. Oxford University Press.
13. Rizzo, A., & Koenig, S. T. (2017). Is clinical virtual reality ready for primetime? Neuropsychology, 31(8), 877–899.
14. Steinberg, L. (2020). Age of Opportunity: Lessons from the New Science of Adolescence(2nd ed.). Houghton Mifflin Harcourt.
15. Taslitz, A. E. (2021). Algorithmic evidence and the future of criminal procedure. *Harvard Civil Rights-Civil Liberties Law Review, 56*(1), 1-48.
16. Tulving, E. (1983). Elements of Episodic Memory. Oxford University Press.
17. Vrij, A., Fisher, R. P., & Blank, H. (2017). A cognitive approach to lie detection: A meta-analysis. Legal and Criminological Psychology, 22(1), 1-21.
18. Završnik, A. (2020). Algorithmic justice: Algorithms and big data in criminal justice settings. European Journal of Criminology, 18(5), 623-642.
19. Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs.