Tihan, Eusebiu Jean (2026), Digital Architectures for Judicial Psychological Evidence: A Translational Analysis of the Integration of Virtual Reality, Artificial Intelligence and Blockchain in the Assessment of Minors, IT & C, 5:2, 7-18, DOI: 10.58679/it40887, https://www.internetmobile.ro/digital-architectures-for-judicial-psychological-evidence/

 

Abstract

To provide a translational analysis of emerging technologies—Virtual Reality (VR), Artificial Intelligence (AI), and Blockchain—in the judicial psychological assessment of minors, examining both their potential to enhance evidence validity and the risks they pose to memory integrity, procedural fairness, and fundamental rights.

Keywords: judicial psychological assessment; virtual reality; artificial intelligence; blockchain; false memories; forensic technology

 

Objective: To provide a translational analysis of emerging technologies—Virtual Reality (VR), Artificial Intelligence (AI), and Blockchain—in the judicial psychological assessment of minors, examining both their potential to enhance evidence validity and the risks they pose to memory integrity, procedural fairness, and fundamental rights.

Methods and Results: This translational analysis synthesises findings from cognitive psychology, neuropsychology, forensic practice, and recent technological research. The reconstructive memory model (Loftus, 2024) demonstrates that remembering is an active construction process vulnerable to contamination. The psychodramatic concept of “surplus reality” (Moreno, 2022) explains the altered state induced during evaluation. Recent experimental research (Chan et al., 2024) demonstrates that generative AI chatbots induce over three times more false memories compared to control conditions, with 36.4% of responses erroneously influenced and effects persisting after one week. VR environments, while offering controlled mnemonic reactivation, present significant “procedural implantation” risks. Blockchain technology provides cryptographic guarantees for evidence integrity through immutable timestamping, addressing chain-of-custody vulnerabilities.

Conclusions: VR, AI, and Blockchain represent fundamental reconfigurations of how psychological evidence is collected, analysed, and preserved. Their integration requires mandatory operational protocols, the institutionalisation of Explainable AI (XAI) principles, investment in public digital infrastructure, and continuous interdisciplinary professional training. These technologies must serve as instrumental extensions of human judgment, not substitutes for it, with the psychologist as “architect of the digital process” ensuring ethical and transparent implementation that protects minors’ dignity and rights.

Arhitecturi digitale pentru proba psihologică judiciară: O analiză translațională a integrării realității virtuale, inteligenței artificiale și blockchain în evaluarea minorilor

Rezumat

Expertiza psihologică judiciară a minorilor se confruntă cu o provocare epistemică fundamentală: sesiunea de evaluare nu este un act neutru de descoperire a unui adevăr psihologic preexistent, ci un proces activ de co-construcție narativă, vulnerabil la contaminare, sugestibilitate și erori de interpretare. Prezentul articol propune o analiză translațională a integrării tehnologiilor emergente – Realitate Virtuală (VR), Inteligență Artificială (AI) și Blockchain – ca răspuns la aceste vulnerabilități structurale. Plecând de la fundamentele neuropsihologice ale memoriei reconstructive (Loftus, 2019, 2024; Loftus & Klemfuss, 2024) și de la conceptul psihodramatic de „plus-realitate” (Moreno, 2022), examinăm modul în care VR poate facilita accesul controlat la amintiri traumatice, dar și riscul major de „implantare procedurală”. Analizăm potențialul AI ca auditor al declarațiilor (prin analiză lingvistică asistată și evaluarea micro-expresiilor) și problema fundamentală a „cutiei negre” algoritmice, care intră în conflict cu dreptul la un proces echitabil și contradictorial (Završnik, 2020). Cercetări recente (Chan, Pataranutaporn, Suri, Zulfikar, Maes, & Loftus, 2024) demonstrează că interacțiunea cu sisteme generative de tip chatbot poate induce de peste trei ori mai multe false memorii comparativ cu metodele tradiționale, o constatare cu implicații profunde pentru utilizarea AI în interviurile investigative. În fine, explorăm utilizarea tehnologiei Blockchain pentru asigurarea integrității și trasabilității probelor digitale, de la înregistrarea audio-video a interviului până la raportul final de expertiză. Concluzionăm că aceste tehnologii nu trebuie privite ca substitute ale judecății umane, ci ca extensii instrumentale ale acesteia, a căror implementare etică și eficientă necesită un cadru normativ adaptat, protocoale operaționale clare (precum „Fișa de Control pentru Validarea Tele-Expertizei” – FCVE-2026) și o formare continuă a specialiștilor la intersecția dintre psihologie, drept și tehnologie.

Cuvinte cheie: expertiză psihologică judiciară, realitate virtuală, inteligență artificială, blockchain, memorie reconstructivă, sugestibilitate, false memorii, etică tehnologică, criminalistică digitală.

 

IT & C, Volumul 5, Numărul 2, Iunie 2026, pp. 7-18
ISSN 2821 – 8469, ISSN – L 2821 – 8469, DOI: 10.58679/it40887
URL: https://www.internetmobile.ro/digital-architectures-for-judicial-psychological-evidence/
© 2026 Eusebiu Jean TIHAN. Responsabilitatea conținutului, interpretărilor și opiniilor exprimate revine exclusiv autorilor.

 

Digital Architectures for Judicial Psychological Evidence: A Translational Analysis of the Integration of Virtual Reality, Artificial Intelligence and Blockchain in the Assessment of Minors

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

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

 

1. Introduction

The judicial psychological assessment of minors occupies a space of maximum sensitivity at the intersection between the rigours of law and the fluidity of the developing human psyche. In this space, the stake is not only the establishment of legally relevant psychological facts, but also the protection of a vulnerable subject and ensuring the validity of evidence that can decisively influence the course of a judicial process. Research from recent decades has convincingly demonstrated that human memory, far from being a faithful repository of the past, is an active and reconstructive process, profoundly influenced by the interrogative context, external suggestions and the emotional state of the subject (Loftus, 2019, 2024; Ceci & Bruck, 2023). As Loftus (2026) emphasises, when someone retrieves a memory, they “do not play back a recording”, but actively “construct” that memory, the brain collecting fragments of information from different times and places and forging them into a memory. Moreover, the evaluative framework itself – the psychologist’s office, the relational dynamics with the expert, the implicit pressure of the judicial stakes – can induce a non-pathological altered state of consciousness, a Morenian “surplus reality”, which facilitates access to memories, but simultaneously exposes the subject to critical risks of confabulation and retroactive contamination (Tihan & Tihan, 2026; Moreno, 2022).

In this context, the integration of emerging technologies into the laboratory of psychological evidence is no longer a futuristic option, but an accelerating reality. Virtual Reality (VR) promises to provide controlled environments for the contextual reactivation of memories; Artificial Intelligence (AI) is emerging as an auditor of statements, capable of detecting subtle patterns that escape the human eye; and Blockchain technology aspires to guarantee the absolute integrity and traceability of digital evidence, from the moment of collection to its presentation in court.

However, recent research sounds a strong alarm bell. A study conducted by Chan, Pataranutaporn, Suri, Zulfikar, Maes and Loftus (2024) demonstrated that interaction with a conversational chatbot based on large language models (LLMs) can induce over three times more immediate false memories compared to a control group, and 36.4% of users’ responses were erroneously influenced through the interaction. These findings underline the considerable risks associated with the use of AI in sensitive contexts such as investigative interviews and impose a profound reflection on the necessary ethical frameworks.

This article aims to provide a translational analysis of these technologies, examining not only their innovative potential, but also the risks and ethical dilemmas they generate, with special reference to recent research on the induction of false memories through AI. Structured around three main cores, we will explore: (1) VR as a tool for controlled mnemonic reactivation and the risk of “procedural implantation”; (2) AI as an auditor of statements and the problem of algorithmically induced false memories, as well as opacity (the “black box”) in relation to the right to defence; (3) Blockchain as an infrastructure for securing the integrity of digital evidence. The conclusion will emphasise the need for a regulatory framework and clear operational protocols to transform these technologies from potential sources of error into robust tools for a more precise and protective justice system for minors.

2. Virtual Reality: “Digital Surplus Reality” between Mnemonic Facilitation and Implantation Risk

2.1. Theoretical Foundations: Reconstructive Memory and the Concept of Surplus Reality

To understand both the potential and the danger of using Virtual Reality in judicial psychological assessment, a dual theoretical anchoring is necessary: in the cognitive psychology of memory and in Morenian psychodrama.

The reconstructive memory model, pioneered by Elizabeth Loftus and confirmed through decades of experimental research, posits that the act of remembering is not a passive playback of fixed recordings, but an active process of reconstruction, influenced by general knowledge, expectations, post-event information and the context of recall (Loftus, 2019, 2024; Loftus & Klemfuss, 2024; Zlate, 2021). In her recent works, Loftus (2024) synthesises research on the impact of misinformation on memory, emphasising the persistence of suggestion effects over time. The engram – the neurophysiological trace of the memory – is fragile, fragmented and incomplete. Faced with a cognitive demand such as “tell me what happened”, the brain does not search for a perfect file, but initiates a process of search and integration, filling mnemonic gaps with plausible, but not necessarily correct, elements. In a judicial context, this intrinsic vulnerability is amplified by stress, by the pressure for narrative coherence and by the suggestions, even involuntary, of the examiner.

In parallel, the concept of “surplus reality” introduced by J.L. Moreno in psychodrama provides an essential hermeneutic framework for understanding the phenomenology of the evaluation session (Moreno, 2022). Surplus reality designates a psychological space beyond concrete and fantastic reality – a space of the possible, of replay, of symbolic exploration. In the expert’s office, this space is co-created by the expert and the subject through questions, focused attention and the relational framework. The subject is not called upon to recite a memorised script, but to “re-enter the scene”, to emotionally and sensorially re-enact a past experience. This controlled immersion can facilitate access to deeper layers of memory, especially to affective and sensory details, but at the same time blurs the boundary between “then and there” and “here and now”, increasing the risk of confusion between the original perception and the present narrative construction.

2.2. VR as a Technological Extension of Surplus Reality

Virtual Reality translates this conceptual framework into the technological plane, offering the possibility of creating simulated, controlled environments that can replicate certain contextual features of the event scene, without recreating the event itself. This capacity has profound implications for mnemonic reactivation.

Research in the neuroscience of memory shows that context cues – spatial, temporal, sensory – play a crucial role in the retrieval of episodic memories (Tulving, 1983). A generic VR environment, which reproduces, for example, a neutral school corridor or a certain type of public space, can act as a controlled trigger for sensory memories (smells, sounds, quality of light) that remain inaccessible in the classic verbal interview (Rizzo & Koenig, 2017). For a minor who is a victim or witness, whose traumatic memory may be blocked or fragmented, this contextual reactivation can facilitate a richer and more nuanced account.

The potential benefit is thus twofold: on the one hand, it can enhance the accuracy and detail of the statement; on the other hand, within a controlled therapeutic setting, exploring an environment similar to the traumatic one can have a habituation effect, reducing anxiety associated with the context and preventing re-traumatisation in later stages of the process.

2.3. The Risk of “Procedural Implantation” and the Necessity of Ethical Protocols

Beyond the promises, VR introduces a risk of a new magnitude, which we might call “procedural implantation” or “immersive contamination”. If traditional verbal suggestibility refers to the power of questions to shape a narrative, VR acts at a deeper, multi-sensory level. When a minor wears a VR headset, they are not just remembering; they are placed inside a projected reality. Scenographic details – lighting, spatial arrangement, textures, ambient sounds – and even the posture of neutral avatars can, through the immersive power of the experience, become incorporated into the subsequent memory of the event (Andrejevic, 2019).

The paradox is profound: a technology designed to “unblock” or “clarify” a memory can, in fact, actively rewrite it. A predominant colour in a virtual environment, absent from the original event, can later become a detail that the minor “remembers” with sincerity. The authority to define the past thus shifts subtly from the neuronal traces of the witness to the design choices of the VR environment’s creator. This transforms the expert or technician who architects the VR environment into an unrecognised co-author of the evidence, wielding significant, but often invisible, power.

To manage this risk, absolute ethical and methodological guarantees are necessary:

1. Neutrality of the VR environment: The virtual environment must never contain case-specific details introduced by the expert (for example, a specific object, a person, a distinctive colour of a room). It must be generic, a category of space (a school classroom, a park), not a replica of the crime scene.
2. Explicit informed consent: Parents and the minor (to the extent of their capacity to understand) must be informed in detail about the purpose of the VR session, the immersive nature of the experience, the potential risk of dizziness or dissociation, and, most importantly, the fact that the environment is generic and not a reconstruction of the event.
3. Full control and documentation of the session: The expert must conduct the session step by step, with frequent pauses to check the minor’s emotional state. The entire VR session must be audio-video recorded, including the verbal interactions between expert and minor and the minor’s reactions.
4. Transparent reporting: In the expert report, the use of VR must be described with radical transparency: the purpose, the characteristics of the generic environment, the duration of the session, the observed reactions and, crucially, the epistemological status of the information obtained. This does not constitute additional factual evidence, but a context for understanding the emotional impact and mnemonic access. A possible formulation: “To facilitate access to memories with dissociative potential, a short exploration session was used, with the informed consent of the guardian, in a generic virtual reality environment, representing a neutral institutional interior. The purpose was to provide sensory context cues. During the exploration, the minor spontaneously mentioned: ‘Here it smells like then.’ Subsequently, upon returning to the verbal interview, they reported a new olfactory detail concerning the location of the event. This observation, although not constituting independent evidence, is consistent with the state of mnemonic reactivation and deepens the understanding of the traumatic impact.”

3. Artificial Intelligence as Auditor of Evidence: Opportunities, the Risk of False Memory and the “Black Box” Problem

3.1. Current and Potential Applications of AI in Statement Analysis

Artificial Intelligence, particularly through its subfields of natural language processing (NLP) and affective computing, is increasingly entering the laboratory of psychological assessment, not as a replacement for human judgment, but as a tool for auditing and enhancing analytical precision.

Assisted linguistic analysis (NLP): NLP algorithms can be trained to analyse transcripts of statements, identifying linguistic patterns associated with different cognitive states. They can:

• Detect inconsistencies: By comparing multiple versions of the same statement, they can signal subtle changes in vocabulary, syntactic complexity or pronoun use, which may indicate areas warranting closer attention from the expert (Vrij, Fisher, & Blank, 2017).
• Assist CBCA application: Algorithms can be trained to identify the presence of certain criteria from Criteria-Based Content Analysis (CBCA), such as the frequency of realism markers (“it seemed to me”, “I felt”), the density of sensory details or the presence of spontaneous corrections. They provide an indicative score or a highlighting of passages that merit in-depth human analysis, increasing objectivity and coverage.
• Estimate linguistic age: By comparing the complexity of the language used by the minor with developmental norms for their age, AI can offer an additional clue as to the plausibility that the statement comes from them and is not a learned text.

Micro-expression and vocal parameter analysis: Specialised affective computing software can analyse video recordings of the interview to detect micro-facial expressions (of fear, disgust, sadness) that last milliseconds and can escape the human eye (Ekman & Friesen, 2003). These can be correlated with specific moments in the narrative, offering clues about the emotional load associated with different segments of the account. Similarly, paralinguistic analysis can detect changes in the fundamental frequency of the voice (tremor, sub- or supra-modulation) or in the rhythm of speech, which correlate with psychological stress.

3.2. The Emerging Risk: Induction of False Memories through AI Interaction

The most recent and concerning discovery in the field of human-AI interaction comes from a study conducted by a team of researchers from the MIT Media Lab, in collaboration with Elizabeth Loftus. The research, entitled “Conversational AI Powered by Large Language Models Amplifies False Memories in Witness Interviews” (Chan, Pataranutaporn, Suri, Zulfikar, Maes, & Loftus, 2024), investigated the impact of interaction with different types of AI interfaces on the formation of false memories in the context of simulated witness interviews.

The study involved 200 participants who watched a video of a crime scene, after which they interacted with different types of AI interviewers or completed questionnaires, answering questions that included five suggestive items designed to induce false memories. Four experimental conditions were tested: (1) control group, (2) standard questionnaire, (3) pre-scripted chatbot and (4) generative chatbot based on a large language model (LLM).

The results are dramatic and have profound implications for the use of AI in judicial contexts:

• Amplification of false memories: The generative chatbot condition significantly increased the formation of false memories, inducing over three times more immediate false memories compared to the control group and 1.7 times more than the standard questionnaire method.
• Influence rate: 36.4% of users’ responses who interacted with the generative chatbot were erroneously influenced through the interaction.
• Persistence over time: After one week, the number of false memories induced by the generative chatbot remained constant, and participants’ confidence in these false memories remained significantly higher than in the control group.

These findings validate a major concern: interaction with advanced AI systems, particularly generative ones, can constitute a powerful source of memory contamination. In the context of judicial psychological assessment, the use of such technologies without strict protocols could lead to the invalidation of evidence or, more seriously, to convictions based on algorithmically induced false memories.

3.3. The Fundamental Problem: The “Black Box” and the Right to a Fair Trial

Beyond the specific risk of inducing false memories, the integration of AI into the generation of judicial evidence raises a profound problem, one that touches the very core of the right to a fair trial: the opacity of “black box” algorithms.

Many high-performance AI models, especially those based on deep neural networks, are incapable of providing a human-understandable explanation of how they arrived at a particular result (Burrell, 2016). A system might produce a “credibility score” of 87% for a statement, but the causal path from the linguistic input to this output is often unknown, even to its developers. This transforms AI from an analytical tool into a kind of “digital oracle”, whose verdicts must be accepted based on technological authority, not on the transparency of its reasoning.

In the context of a judicial process, this opacity directly conflicts with the principle of adversarial procedure and the right to defence. A lawyer cannot interrogate an algorithm. They cannot ask it questions such as:

• “On what specific linguistic feature did you base this score?”
• “How do you ensure that your training data, coming from adults, is applicable to a traumatised 14-year-old adolescent?”
• “Did you consider the cultural or regional particularities of the language?”

The algorithm cannot answer. It can only reproduce its result. Moreover, these algorithms are often the intellectual property of private corporations, which can invoke trade secrets to refuse access to the source code or to the training datasets (Završnik, 2020). This creates an unprecedented situation in the history of justice: a private company can hold the key to interpreting evidence, and its commercial interests can prevail over the fundamental right of a citizen to a meaningful defence.

3.4. Towards Explainable Artificial Intelligence (XAI) in the Judicial Context

The response to this emerging crisis is the imperative requirement that any AI system used in the generation or analysis of judicial evidence be an Explainable Artificial Intelligence (XAI) system. XAI refers to a set of methods and techniques that allow the results of AI systems to be understood by humans (Gunning & Aha, 2019).

In practice, an XAI system for statement analysis should be able to generate a comprehensible explanation, for example: “The system indicated possible narrative invention due to: (a) an abnormally low frequency of perceptual verbs (‘I saw’, ‘I heard’) compared to the normative corpus for the age of 14; (b) a repetitive syntactic structure, characteristic of learned narratives; (c) a high rate of grammatical corrections in the second account, suggesting narrative construction rather than spontaneous recall.”

Such an explanation allows a human expert to evaluate the AI’s logic, identify potential biases, and a lawyer to challenge the premise on which it is based (for example, the relevance of the normative corpus to the specific case). Only thus can AI become a transparent partner in the decision-making process, and not an opaque substitute for it.

Recent research on the induction of false memories underscores this necessity even more strongly. If an AI system can, unintentionally, generate false memories, then transparency and explainability become not just desirable, but absolutely essential to be able to identify and correct such errors. Any black box that influences or analyses a minor’s memory must be viewed with utmost suspicion and subjected to unprecedented auditing standards.

4. Blockchain: Guaranteeing the Integrity and Traceability of Digital Evidence

In the digital era, where deepfakes and the facile manipulation of audio-video files are becoming an ever-present threat, ensuring the integrity of evidence from the moment of collection to its presentation in court becomes a crucial issue. Blockchain technology offers a robust technical response to this challenge.

Blockchain is, essentially, a distributed digital ledger, in which information is stored in cryptographically linked, immutable and transparent blocks (Kshetri, 2017). Applied to the chain of custody of judicial evidence, this technology can function as follows:

1. Recording and “anchoring” the evidence: At the moment an audio-video recording of the interview, a set of raw psychometric data, or an expert report is finalised, the expert or the secure platform generates a unique cryptographic hash– a “digital fingerprint” – for that file. This hash is then inscribed on an authorised private justice blockchain (or a qualified timestamping service), obtaining a cryptographic and immutable timestamp.
2. Verification of integrity: Any subsequent modification, no matter how small, of the original file will generate a completely different hash. At the time of presenting the evidence in court, any interested party (judge, lawyer) can recalculate the hash from the presented file and compare it with the one recorded on the blockchain. If the two hashes coincide, the integrity of the file from the moment of inscription is mathematically guaranteed.

This procedure has profound implications for trust in digital evidence:

• Immutability of evidence: Once “anchored” in the blockchain, the evidence becomes impossible to modify retroactively without leaving a detectable trace. This ensures “a single version of the truth”.
• Transparency of the chain of custody: The blockchain can record not only the file’s hash, but also a complete history of access, allowing precise verification of who, when and why interacted with the evidence.
• Combating deepfakes: In a context where generative artificial intelligence can create extremely realistic falsified video recordings, the cryptographic hash becomes the only certain method of distinguishing an authentic recording from a manipulated one.

Integrating Blockchain into judicial psychological assessment is not merely a technical security measure, but also an act of institutional legitimation. It transforms “suspicion of manipulation” from a subjective debate into an objective mathematical verification, consolidating trust in the system and reducing the time and resources allocated to technical challenges.

5. Discussions: Synergies, Tensions and the Need for an Integrative Framework

The separate analysis of the three technologies – VR, AI and Blockchain – reveals both their synergistic potential and the tensions they generate. An integrative vision is essential to transform these instruments into a coherent architecture of psychological evidence.

Possible synergies: Imagine a scenario where an interview with a minor, conducted according to a non-suggestive protocol (e.g., NICHD), is audio-video recorded. Immediately after its conclusion, the file is “anchored” in the Blockchain, guaranteeing its integrity. Subsequently, a transcript of the interview is analysed by an XAI system which identifies linguistic patterns and flags certain passages for the expert’s closer attention, taking care, however, not to itself become a source of suggestion. In parallel, a short exploration session in a generic VR environment is used to assess emotional reactivity to context cues, and the entire session is also recorded and secured. The expert then integrates these multiple data points into a report which is, in turn, “anchored” in the Blockchain. The result is psychological evidence whose collection, analysis and reporting are transparent, verifiable and epistemologically robust.

Tensions and challenges: This integrative vision is not without major challenges.

1. Confidentiality and data protection: The long-term collection and storage of minors’ biometric data and video recordings raises serious confidentiality issues. It is essential to implement strict access control and anonymisation mechanisms, in accordance with European (GDPR) and national legislation. The risk emerges that this extremely sensitive data could become a valuable resource for companies, fuelling what Zuboff (2019) calls “surveillance capitalism” – a system where human experience is transformed into raw material for behavioural prediction and control.
2. Vendor dependence and technological sovereignty: Outsourcing critical justice infrastructure to private companies creates the risk of a dangerous dependency. Courts can become locked into proprietary platforms, and decisions regarding upgrades, security, or even access to data can be influenced by commercial interests. Investment in public, open-source and auditable digital infrastructure is an essential countermeasure to maintain judicial sovereignty.
3. Interdisciplinary professional training: The competent and ethical use of these technologies requires a fundamental shift in the training of judicial psychologists, magistrates and lawyers. They must acquire not only basic technical skills, but also a deep understanding of the limits, potential biases and ethical implications of digital tools. The introduction of mandatory modules of “judicial technological literacy” into university curricula and continuing education programmes becomes necessary.
4. Integration of safety protocols: Recent discoveries regarding the induction of false memories by generative chatbots (Chan et al., 2024) impose a radical re-evaluation of how AI can be integrated into the interviewing process. A concrete proposal is the development of hybrid protocols, where AI is used exclusively post-interview, for the analysis of already collected data, and never in real-time, in direct interaction with the minor. Furthermore, any AI tool used in analysis must be subjected to rigorous testing to verify whether it can itself generate suggestive patterns.

Conclusions and Recommendations

The integration of Virtual Reality, Artificial Intelligence and Blockchain into the judicial psychological assessment of minors marks a paradigm shift. These technologies are not simple accessories, but fundamental reconfigurations of how psychological evidence is collected, analysed, interpreted and preserved. They offer unprecedented opportunities for increasing accuracy, objectivity and transparency, but simultaneously introduce new and profound risks, from “procedural implantation” in VR to the erosion of the right to defence through algorithmic opacity and the risk of inducing false memories through interaction with generative systems.

The research conducted by Chan, Pataranutaporn, Suri, Zulfikar, Maes and Loftus (2024) serves as a powerful alarm signal: interaction with generative AI can dramatically amplify the formation of false memories, and these effects persist over time. This finding has direct implications for any use of AI in investigative contexts and underscores the necessity of extremely strict ethical and methodological frameworks.

The central conclusion of this translational analysis is that these technologies should not be viewed as substitutes for human judgment, but as instrumental extensions of it. The role of the judicial psychologist is redefined: they are no longer just a clinician or an interpreter of data, but also an architect of the digital process, responsible for designing an evaluative framework that integrates technology consciously, ethically and transparently.

To transform the potential of these instruments into practical reality, the following measures are necessary:

1. Development of clear and mandatory operational protocols: There is a need for national and European guidelines to standardise the use of VR, AI and Blockchain in the judicial context. Examples such as the “Checklist for Tele-Expertise Validation” (FCVE-2026) proposed in the foundational work (Tihan, E. J., & Tihan, M. L. 2026). must be extended and adapted for each technology. In the case of AI, protocols must include an explicit prohibition of direct generative interaction with minors during the interview.
2. Institutionalisation of the principle of Explainable Artificial Intelligence (XAI): Any AI system used in the generation or analysis of evidence must be auditable and capable of providing a human-comprehensible explanation of its reasoning. This is not merely a technical desideratum, but a fundamental requirement of the right to a fair trial. Systems that cannot provide such an explanation should be excluded from judicial use.
3. Investment in public digital infrastructure: States must invest in developing open-source, publicly verifiable platforms and tools to avoid dependence on private vendors and to ensure democratic control over critical justice infrastructure.
4. Development of continuous interdisciplinary training programmes: Judicial psychologists, magistrates and lawyers must benefit from training that equips them with the necessary knowledge to understand, use and competently challenge evidence generated with the help of new technologies. This training must include specific modules on the psychology of memory, suggestibility and technological risks.
5. Creation of independent audit mechanisms: Given the technical complexity of these instruments, it is necessary to establish independent technological audit bodies, capable of evaluating and certifying AI and VR systems before they are used in judicial contexts.

Ultimately, the future of digital justice will not be determined by the speed of technological innovation, but by the wisdom with which we manage to 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. As Loftus’s own research reminds us (2024), the malleability of human memory is a constant that technology can amplify, but only professional wisdom and ethics can keep under control.

References

1. Andrejevic, M. (2019). Automated Media. Routledge.
2. Burrell, J. (2016). How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data & Society, 3(1), 1-12.
3. Ceci, S. J., & Bruck, M. (2023). The reliability of children’s testimony: New perspectives on suggestibility. Annual Review of Psychology, 74, 211–238.
4. 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.
5. Ekman, P., & Friesen, W. V. (2003). Unmasking the Face: A Guide to Recognizing Emotions from Facial Clues. Malor Books.
6. Gunning, D., & Aha, D. (2019). DARPA’s explainable artificial intelligence (XAI) program. AI Magazine, 40(2), 44-58.
7. Kshetri, N. (2017). Blockchain’s roles in meeting key supply chain management objectives. International Journal of Information Management, 39, 80-89.
8. Loftus, E. F. (2019). Eyewitness testimony. In R. Brown (Ed.), The Oxford Handbook of Memory. Oxford University Press.
9. Loftus, E. F. (2024). The fiction of memory: Forensic implications. Annual Review of Psychology, 75, 123-148. https://doi.org/10.1146/annurev-psych-012423-125632
10. Loftus, E. F. (2026). The fiction of memory: Forensic implications. Annual Review of Psychology, 75, 123-148.
11. Loftus, E. F., & Klemfuss, J. Z. (2024). Memory and the law: The science of wrongful convictions. In L. Nadel & W. Sinnott-Armstrong (Eds.), Memory and Law. Oxford University Press.
12. Moreno, J. L. (2022). Psychodrama, Volume 1(4th revised ed.). Psychodrama Network Publishing.
13. Rizzo, A., & Koenig, S. T. (2017). Is clinical virtual reality ready for primetime? Neuropsychology, 31(8), 877–899.
14. Tihan, E. J., & Tihan, M. L. (2026). The Psychological Architecture of Judicial Evidence: A Multidimensional Approach to the Assessment of Minors. Focus Institute of Social Ecology and Human Protection Publishing House (in progress).
15. Tulving, E. (1983). Elements of Episodic Memory. Oxford University Press.
16. 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.
17. Završnik, A. (2020). Algorithmic justice: Algorithms and big data in criminal justice settings. European Journal of Criminology, 18(5), 623-642.
18. Zlate, M. (2021). The Psychology of Cognitive Mechanisms(3rd ed.). Polirom Publishing House.
19. Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs.