Integrating Generative AI in Educational Psychology Assessments: Preliminary Ethical and Practical Guidelines

Abstract

This article presents an examination of the intersection between generative AI and psycho-educational assessments, offering a comprehensive overview of their integration. It begins by exploring the fundamentals of artificial intelligence, emphasizing the role of generative AI in Natural Language Processing (NLP) and its potential implications for psychological research and clinical practices. Concurrently, the manuscript delves into the intricate steps of a psycho-educational assessment for children, elucidating the multifaceted process encompassing cognitive, emotional, and academic evaluations. Through this integrated exploration, the article not only provides detailed guidelines, an ethical framework, and nuanced insights into potential challenges but also addresses the global shortage of trained clinicians. By examining how advanced technologies including Generative AI can assist in enhancing diagnostic efficiency, when outputs are interpreted and validated by trained professionals. Drawing on the perspectives highlighted by Duff and Roberts (2016) and insights from generative AI, the article proposes guidelines that streamline assessments, prioritize ethical considerations, and maximize the benefits of human expertise.

Introduction

Educational psychology experiences a critical shortage of qualified clinicians, as highlighted by Moody (2025), underscoring the need for innovative solutions to optimize and expedite the assessment process. Artificial Intelligence (AI) emerges as a promising technology that holds the potential to revolutionize how assessments are conducted. Thus AI and related technologies, by supporting scoring and reporting, offer a potential suite of improvements encompassing efficiency, accuracy, reliability, and the standardization of assessment tools, as articulated by Johnson et al. (2020). Specifically, the application of AI and related technologies in educational assessments is not only a pragmatic response to the scarcity of clinicians, but may also lead to a transformative shift in the methodology itself.

Building upon the insights presented by Duff and Roberts (2016), who observed a seismic shift in the lives of children due to technological advances, this manuscript embarks on an exploration of the integration of specifically generative AI and related technologies in educational assessments. The purpose of which is twofold: (1) to provide practitioners with a balanced and critical understanding of opportunities and limitations presented by AI, and (2) to ensure practitioners are aware of the requisite components for AI to be responsibly designed for research, education and clinical settings. The broader purpose extends beyond the aforementioned pragmatic goals to encompass ethical considerations, preservation of nuanced understanding of human cognition and behavior, and the interaction between AI design and human expertise, thus providing valuable insights that inform both practitioners and researchers, fostering a balanced and informed perspective on the evolving landscape of AI in educational psychology.

A Basic Overview of Current Forms of AI

AI encompasses a variety of technologies that enable machines to perform tasks that typically require human intelligence, thus allowing machines to mimic aspects of human reasoning. Within AI, Natural Language Processing (NLP) stands out as a crucial component, focusing on the interaction between computers and human language. NLP allows machines to analyze, process, and generate text based on statistical patterns in human language, thus allowing machines to generate human-like language, facilitating interaction between computers and users.

Within NLP, generative AI represents a specialized subtype that involves the creation of new content, such as text or dialogue, mimicking human-like language generation. Another facet of AI is Computer Vision (CV), which involves teaching machines to process visual information from the world, simulating human visual perception. This enables AI systems to analyze and interpret images, videos, and other visual data. Speech processing extends this capability to the auditory domain, enabling machines to interpret, analyze, and generate audio signals. These elements form the foundation of AI technologies, presenting psychologists with a diverse set of tools and challenges.

Supervised and unsupervised learning are foundational approaches within AI. Supervised learning involves training AI algorithms with a labelled dataset, where each example is annotated or coded by humans. This meticulous guidance enables the algorithm to learn patterns and associations, making it adept at making predictions or classifications on new, unseen data. The drawback of supervised learning lies in its reliance on data annotations, a process that is both time-consuming and costly. Despite the associated drawback, this is the preferred method by many psychologists for its transparency and decreased implicit bias. Nevertheless, data annotation often constrains development of supervised models to smaller datasets relative to those involved in unsupervised models, thus posing a potential bottleneck to achieving high accuracy or success in highly nuanced task fulfilment. Importantly, supervised learning can offer transparency in decision-making, although it does not inherently eliminate bias and may reproduce biases present in labelled data. Unsupervised learning takes a different approach by utilizing unlabeled data. In this method, the algorithm explores patterns and relationships within the data without explicit human guidance. This approach is particularly valuable in handling vast amounts of unannotated information available on the internet, allowing AI systems to uncover inherent structures and associations autonomously. The unsupervised approach is especially relevant for generative AI when dealing with diverse and continuously evolving outputs (e.g., summarization as opposed to text classification), providing flexibility and adaptability. However, unsupervised models are unable to achieve high accuracy when a preponderance of data does not exist for a given population. This is particularly relevant for educational psychology, as data on K-12 student populations is limited. Further, for marginalized K-12 students, such as students with learning disabilities or extra language learners, data is scarce.

Both approaches have limitations and opportunities (Bdaiwi et al., 2025). While supervised learning is constrained by the availability of annotated data, it can provide transparency and accountability when applied to educational decision-making. Unsupervised methods allow flexibility and adaptability but risk embedding latent bias if datasets are unrepresentative. For psychologists, understanding the interplay of these approaches is crucial to evaluating claims about AI’s role in assessment. Notably, generative AI models can also produce content that appears coherent and plausible but may contain inaccuracies or “hallucinations” information that is not grounded in the underlying data. This highlights the necessity of professional oversight when interpreting AI-generated outputs. Further, noise in both supervised and unsupervised AI models highlights an additional reason for the necessity of professional oversight. Noise refers to random or irrelevant variations in data that obscure the underlying patterns the model is intended to learn (Kahneman et al., 2021). Additionally, variations due to noise are often more difficult to detect than bias, as bias often trends in one direction, whereas noise causes variability (Nettleton et al., 2010). Within supervised models, noise appears due to inconstant labelling of data or corrupted input data. Within unsupervised models, noise appears due to variation in input data, as these models rely on the implicit structure within datasets.

In summary, supervised learning uses labelled datasets, whereas unsupervised learning identifies patterns in unlabeled data. Supervised learning, with its reliance on labelled datasets, provides a more structured and directed approach, offering finite, well-defined responses based on human-provided annotations. Conversely, unsupervised learning navigates unannotated data to uncover patterns and associations autonomously. This duality highlights the versatility of AI in handling both well-structured tasks and the exploration of uncharted information landscapes. Understanding the interplay between these learning methods is crucial for psychologists as they explore the evolving dynamics of AI and its impact on human cognitive functioning, emphasizing the need to navigate the spectrum of finite and infinite responses within the context of AI.

Exploring the Intersection of AI and Psychological Practice: Opportunities and Challenges

NLP has the greatest immediate relevance for assessment and intervention, since current methods are mainly language-based interactions, either by text or oral conversation, such as student interviews, intake questionnaires, behavioral observations, and psychometric tests. Thus, an NLP-powered system could generate preliminary summaries of a student’s self-report interview, which the psychologist verifies for accuracy and clinical relevance. However, other forms of AI, particularly in the domains of computer vision and speech processing, hold promising potential. Computer vision systems could analyze recorded classroom observations to identify patterns in attention or engagement, while speech processing might detect variations in tone that could correlate with emotional states. These examples illustrate how AI might support, but not replace, professional practice. Importantly, AI tools that suggest patterns in tone or attention via computer vision or speech analysis have not yet been validated for formal assessment and require professional interpretation. For practitioners, the key opportunity is to critically evaluate whether AI tools offer valid and reliable outputs that meaningfully inform decision-making. For developers, the responsibility is to design systems that are transparent about their data sources, limitations, and accuracy claims. Clear communication between these groups is necessary if AI is to be responsibly integrated into psycho-educational practice. Further, practitioners must remain vigilant for AI hallucinations, as summaries, scoring suggestions, or intervention ideas produced by AI must be critically reviewed and validated by trained professionals.

Exploring these avenues offers psychologists an exciting prospect to integrate AI modalities into their research and clinical practices, shaping a more comprehensive and nuanced understanding of human behavior and cognition.

Ethical Considerations for Integrating AI into Real-World Systems

Multiple ethical concerns require sustained attention to ensure the responsible development and use of AI technologies. These include privacy and confidentiality, equity and bias, transparency of decision processes, and accountability for outcomes.

Bias deserves particular focus. Bias is not only embedded in human decision-making but is also present in the text and data AI systems rely on. Practitioners play a crucial role in recognizing and mitigating bias, although the greater responsibility rests with AI developers, who must incorporate interdisciplinary oversight and equity-focused review processes into product design. Further detail on algorithmic bias in education has been investigated by Baker and Hawn (2022) and Schiebinger, et al (2018), providing useful insight into how and why language-based AI models can become biased. To mitigate algorithmic bias, practitioners must be involved in the creation of robust and sensitive frameworks for algorithmic bias detection.

Some ethical issues, such as copyright law, data protection, and cybersecurity, fall largely outside the control of school psychologists. Although this manuscript details technical issues, this is to ensure practitioners are aware of oversights that some AI technologies may have and can, in turn, recommend solutions. This allows practitioners to critically evaluate claims made by AI vendors, demand transparent reporting of model limitations, and advocate through professional organizations for responsible standards. Finally, discussions of AI ethics should avoid anthropomorphising language. AI does not “decide” or “understand”; it generates outputs based on statistical patterns. This framing helps set realistic expectations for what AI can and cannot contribute to psychometric assessment and human behavior.

Key Concerns for Educational Psychologist Adoption of AI Technologies

Privacy concerns: AI systems often rely on vast amounts of data to function effectively. In educational psychology, this may involve personal data about students, learning behaviors, mental health, and emotional states. Practitioners should be able to evaluate whether these protections are described transparently in product claims. Specifically, practitioners should ensure AI developers implement strong data protection mechanisms such as encryption (converting information into a secure format that is unreadable) and data anonymization (removing or altering identifiable information) and data minimization is vital, ensuring that AI only accesses the data necessary to achieve the intended goal, which helps to mitigate the risk of privacy violations.

Copyright concerns: While technical copyright compliance falls largely under the responsibility of developers, the use of copyrighted materials for training AI models presents significant challenges. AI systems in educational psychology often rely on datasets, some of which may include proprietary content or materials protected by copyright. For practitioners, the key responsibility is to be aware that not all AI tools have transparent data sources and to demand clarification when products are marketed for school use. Further, AI systems that generate content autonomously pose risks related to AI-generated content that might inadvertently infringe on existing copyrights. For educational psychology, where content such as learning modules or psychological tools may be generated. Practitioners should critically review AI outputs before adopting them for instruction or assessment.

Addressing biases: Bias in AI can have far-reaching consequences, particularly in educational psychology, where AI systems may make recommendations affecting students’ development. Ensuring data diversity is crucial. AI should be trained on diverse and representative datasets to avoid embedding societal biases into its algorithms (Belenguer, 2022; Ferrara, 2024). This responsibility falls primarily on developers, who must integrate a variety of methods and interdisciplinary oversight into training processes. Further, the detection of biases within AI systems should be an ongoing process involving both practitioners and developers, employing a variety of tools and methodologies to ensure fairness and equity. This is especially relevant as biased AI could disproportionately disadvantage certain groups of students based on race, gender, or socioeconomic background, exacerbating existing inequalities. Fundamentally, responsibility for reducing bias must be shared: developers must design systems that integrate fairness checks, and practitioners should evaluate claims, demand accountability, and remain alert to inequitable outcomes.

Handling unexpected responses: AI systems can behave unpredictably, especially when deployed in human‑centric areas. It is crucial to maintain human oversight to monitor and intervene when AI makes questionable recommendations. AI‑supported assessments should always be reviewed and confirmed by educators or psychologists before use. Proper error‑handling mechanisms should also be built into the system to allow for fallback options if the AI generates inappropriate or harmful outputs. Transparency is key. AI should provide explainable processes rather than being described as “deciding” on its own. This ensures that both students and educators understand how AI reaches its conclusions, building trust and facilitating informed intervention when needed.

Accountability: It is important to clearly define responsibility for the recommendations and actions taken by AI. For example, who is responsible when an AI system causes an incorrect assessment? Should the blame fall on the developers, the users (e.g., educators or psychologists), the institutions that approved it, or perhaps all of them? AI itself should not be described as responsible since it lacks agency; accountability lies with human actors and institutions. Practitioners must ensure that responsibility for any AI‑driven decisions or recommendations is clearly defined and that redress mechanisms are in place for any errors or harms caused. Redress mechanisms are also necessary to ensure that students or teachers can challenge recommendations made by AI if they believe they are unfair or incorrect.

Transparency and explainability: For AI to support assessment of student performance or psychological health, algorithmic transparency is critical. Stakeholders should be able to understand, in clear non‑technical terms, how AI systems come to conclusions, especially in areas where decisions have significant consequences. Explanations for AI decisions help build trust in the system, and training materials for practitioners should emphasize limitations as well as benefits. Transparency efforts should also include highlighting the potential for AI hallucinations, so that stakeholders understand that outputs may contain inaccuracies, and human review is essential for maintaining validity and ethical standards. Further, the decisions AI informs about student assessments or psychological diagnoses must be clear to educators, parents, and students alike. However, AI systems often operate using “black‑box” models, making it difficult to explain why the system reached a particular decision. This lack of transparency can erode trust in the system. Practitioners should insist on clear vendor documentation and avoid adopting tools that cannot provide interpretable outputs.

Compliance: Compliance with regulatory frameworks related to education and privacy is equally important, ensuring that AI systems align with laws protecting students and other vulnerable populations. Practitioners should verify whether products meet the requirements of FIPPA, PIPEDA, or other applicable regulations. Practitioners should also ensure that AI tools conform to provincial standards in addition to federal ones. Ensuring compliance helps avoid legal repercussions, fines, and reputational damage.

Social impact: A social impact assessment (SIA) is a process that evaluates the effects of implementing AI on various aspects of societal well‑being. When considering the integration of AI into education and psychology, SIA ensures that the broader implications are analyzed, not just the immediate operational benefits. Existing frameworks for SIA provide models for how assessments can be conducted, by whom, and what domains should be reviewed. This assessment should address key areas, including: education and learning outcomes; student development; job displacement and workforce implications; equity and access; wider societal norms and values; and ethical considerations. There are few frameworks specific to psychology. Nevertheless, some exist. Skoric et al. (2024) provide a specific framework for SIA, detailing normative questions, process rules, methodology, engagement, and oversight, although their article is tailored for developers, regulators, and standardization initiatives. Fitzgerald and Taylor (2025) offer useful insight into how these frameworks could be developed in a robust and sensitive manner, specifically within psychology. By conducting a comprehensive SIA, stakeholders can make decisions about implementing AI in ways that maximize benefits while minimizing risks and unintended consequences. This proactive approach fosters a balanced integration of technology that aligns with societal values and enhances overall well‑being. Further, the broader social and psychological implications of AI are likely to endure, thus AI has the potential to influence social behavior, educational outcomes, and mental well‑being. For instance, relying heavily on AI for decision‑making may perpetuate reduced human interaction or reliance on automated assessments that overlook the nuanced needs of students. This can lead to unintended consequences, such as undermining students’ confidence or creating over‑reliance on technology. Practitioners must carefully monitor these impacts and use AI only as a supplement to, not a replacement for, professional judgment.

Ethical use of AI: Where AI impacts diagnostic decisions, the ethical development and use of AI is critical. AI systems should have clear purpose limitations, being deployed only in scenarios where they provide ethical and socially beneficial results. Practitioners should ensure they adopt only tools with demonstrated validity and safety.

Continuous monitoring and evaluation: AI must be continuously monitored to ensure it meets ethical standards and functional requirements. Regular audits of AI systems ensure that they behave as expected and that any new ethical concerns are identified and addressed early. Practitioners should ensure that AI tools adopted have feedback mechanisms designed to capture practitioner, student, and parent input, ensuring the system evolves responsibly and remains aligned with ethical expectations and practical outcomes.

Evolving legal and regulatory frameworks: As AI technology continues to evolve, so too will the laws and regulations that govern its use. Practitioners should stay informed and demand documentation from vendors that confirms compliance.

User trust and consent: Obtaining user consent is crucial. In this context, informed consent helps ensure that users, students, or patients understand how their data will be used, stored, and shared, respecting their autonomy and trust. These processes help prevent unauthorized access or misuse, particularly when the data involves vulnerable populations such as children. However, maintaining user trust will be an ongoing challenge. As AI becomes more integrated into psychology, concerns about data use, consent, and privacy are likely to remain. Practitioners should focus on clear communication strategies to ensure students, clients, and parents understand how AI is being used and how they can opt out if they choose.

Lack of Socio-Emotional Nuance

A crucial consideration is the potential risk of integrating algorithmically generated responses into real-world systems. These responses may sometimes be inappropriate or contextually insensitive, introducing significant ethical concerns that require careful human oversight. Despite the remarkable capabilities of today’s generative AI, it does not “understand” socio-emotional nuance but instead predicts likely responses based on patterns in training data. Consequently, there is a risk that AI-produced text may not align with the socio-emotional context of specific situations, potentially leading to misunderstandings, miscommunications, or unintended consequences. Human oversight must remain central to mitigating these risks. For example, the AI Now Institute researches and provides guidelines for ethical AI deployment in various fields, including psychology (Accountable Tech, AI Now Institute, & Electronic Privacy Information Center, 2023). Their work ensures that AI applications in mental health are fair, transparent, and respectful of user privacy. IBM Watson Health implements strict ethical guidelines in its AI applications, ensuring patient data is protected and that AI recommendations are transparent and explainable (International Business Machines Corporation, n.d).

In summary, while AI offers many potential benefits for education and psychology, significant ethical concerns will likely persist. Issues surrounding privacy, bias, transparency, and accountability will continue to challenge AI developers and practitioners. By staying vigilant, adapting to new developments, and prioritizing ethical AI practices, professionals can navigate these challenges while ensuring that AI delivers positive outcomes in critical areas of human development. Further, by addressing ethical considerations, organizations can ensure that their AI systems are developed and deployed responsibly and fairly. Implementing proactive strategies to reduce bias such as interdisciplinary input and regular audits and to assess broader impacts through social impact assessments with clear frameworks is crucial. Strong quality assurance practices and continuous monitoring enhance reliability and trustworthiness.

Cyber Security Considerations for Integrating AI into Real-World Systems

AI systems are inherently vulnerable to attacks, and these vulnerabilities pose significant risks, particularly in fields such as educational psychology, where sensitive personal and psychological data is processed. Security in AI encompasses a range of measures aimed at safeguarding the integrity, availability, and confidentiality of the system and its data. While the technical implementation of these measures is the responsibility of developers, psychologists and educators must be able to ask informed questions about whether tools meet security standards.

Vulnerability management: Vulnerability management involves the continuous identification, assessment, and mitigation of security gaps within an AI system. AI models can develop unexpected vulnerabilities over time. These gaps can be exploited by attackers to manipulate the AI’s behavior or gain unauthorized access to sensitive data. Practitioners should ensure that AI systems under consideration implement these practices or equivalents. Best practices include: Regular Security Audits: Patch Management, Adversarial Testing, Dependency Management.

Access control measures: Strict access control measures are critical to ensuring that only authorized individuals have access to the AI system and personalized data. Sensitive data, such as student performance, psychological assessments, or behavioral trends, must be protected. Key elements of access control include: Role‑Based Access Control (RBAC), Multi‑Factor Authentication (MFA), Logging and Monitoring, Data Segmentation.

Encryption and data protection: Encryption plays a key role in securing data within AI systems, ensuring that even if data is intercepted, it cannot be read without proper decryption keys. Best practices include: Data Encryption in Transit and at Rest and Key Management.

Incident response planning: Even with robust preventative measures, security incidents can still occur, and organizations must be prepared to respond effectively. An incident response plan is a documented, structured approach to handling security breaches, minimizing their impact, and restoring functionality. It should include: Detection and Reporting, Containment and Mitigation, Recovery, and Post‑Incident Analysis.

In educational psychology, security risks are amplified because of the sensitivity of the data involved. A breach could compromise assessments, interventions, or learning outcomes, underscoring the importance of institutional policies that prevent adoption of AI tools lacking independent security review. An attacker could manipulate AI outputs to provide incorrect recommendations, potentially harming students’ learning outcomes or mental health interventions. This highlights the importance of maintaining human oversight of AI-generated recommendations. By implementing these measures, organizations can reduce the risks associated with deploying AI systems, ensuring that the benefits of AI can be realized without compromising data integrity, user trust, or system functionality. For practitioners, the priority is not implementing these technical safeguards directly, but demanding transparency from vendors and understanding where human oversight remains essential. Ethical and security considerations fall between developer responsibility (building secure systems) and practitioner responsibility (evaluating claims and demanding accountability). Addressing these concerns helps ensure that AI contributes positively to society, benefiting users while minimizing risks. Thus, this list provides what practitioners should expect and demand from AI tools within their practice. Further, it ensures practitioners can critically engage with and understand the functions and limitations of cyber security measures.

Exploring the Impact of Generative AI on Professional Psychological Assessment

Transitioning from the broader discussion on the ethical implications of AI, here are specific applications of AI for the psycho-educational assessment for children. This process, designed to comprehensively evaluate cognitive, emotional, and academic functioning, exemplifies the intricate interplay between technology and professional practice. As we navigate through the steps involved in a psycho-educational assessment, we will explore how the integration of generative AI could potentially influence and augment this crucial aspect of psychological evaluation and intervention. By examining this case study, we aim to gain insights into both the opportunities and challenges presented by the utilization of AI technologies in real-world psychological practice. A psycho-educational assessment for a child is a comprehensive process designed to evaluate various aspects of their cognitive, emotional, and academic functioning. The assessment is typically conducted by an interdisciplinary team, including psychologists, educational diagnosticians, and possibly other specialists. Below is an overview of a psycho-educational assessment and how AI could potentially support each step or what would be done if AI was implemented into the assessment process. These are meant to provide examples that are none exhaustive as different diagnoses and psycho-educational assessment may involve different composites of psychometric assessments. However, for any AI-generated recommendation, while useful for highlighting patterns, they may include hallucinations. These outputs cannot be treated as conclusive evidence; human validation is essential before incorporating them into assessment reports or intervention planning. Here is a brief overview of the steps involved in a typical psycho-educational assessment with a more detailed version below. Psycho-educational assessment begins with a referral and initial information gathering, followed by obtaining informed consent. Interviews with parents, teachers, and the child provide crucial insights, complemented by observations in various settings. Standardized cognitive and academic tests, as well as emotional and behavioral assessments, contribute to a comprehensive understanding. Collaboration with specialists and professionals is integral, ensuring a holistic evaluation. Once data is collected, it undergoes rigorous analysis to identify patterns and areas of concern. The findings are then shared with parents, forming the basis for intervention planning and recommendations.

An Overview of the Steps Involved When a Student Undergoes a Psycho-Educational Assessment

Referral

The initial stages of the assessment process typically commence with a referral initiated by parents, teachers, or other concerned individuals who may observe challenges or areas of potential concern in the child’s development or academic performance.

Generative AI could support the process of organizing information across different individuals. However, as informed consent has not been obtained, despite AI ability to perform the task, it is not advisable to be implemented at this stage.

Informed Consent

In the crucial stage of obtaining informed consent, parents or legal guardians are given comprehensive information about the information collection process, assessment process, its objectives, and the specific tests to be administered. This includes a detailed explanation of the purpose behind the assessment, highlighting its significance in understanding the child’s development, identifying areas of strength and need, and guiding appropriate interventions. Additionally, parents receive clarification on the nature of the tests, including their format, duration, and potential implications for the child’s educational journey. This transparent communication ensures that parents have a clear understanding of what to expect during the assessment process and can make informed decisions regarding their child’s participation. To implement AI supporter tools in this process, practitioners must clearly explain how any AI-supported tools will be used in the assessment. This ensures that AI usage does not undermine trust in assessment procedures or practitioners.

Information Gathering

Upon receiving the referral, comprehensive information gathering ensues, aimed at compiling a detailed profile of the child’s developmental history, academic achievements, behavioral patterns, and any specific concerns noted by caregivers or educators. This phase involves gathering data from multiple sources, including parent interviews, teacher observations, previous assessment reports, and relevant medical records. AI could assist this process by summarizing reports or highlighting potential areas of concern across reports, but all outputs must be verified by human practitioners to avoid overreliance and ensure accuracy.

Interviews

Interviews are conducted with parents, teachers, and the child to gather more detailed information about the child’s behavior, social interactions, and academic strengths and challenges. The interview process in psycho-educational assessments follows structured guidelines while allowing for flexibility to adapt to each child’s unique needs and circumstances. Typically, there is a standardized set of questions covering various domains such as cognitive, emotional, and behavioral functioning. However, the flow of the interview may vary based on the child’s responses, non-verbal cues, and the diagnostician’s professional judgment. While some questions may be scripted in advance to ensure consistency across assessments, the interview also allows for fluid conversation to foster rapport and gather comprehensive information. This balance between structured questioning and open dialogue is essential for eliciting meaningful insights into the child’s strengths, challenges, and underlying needs. AI could support this process through organizing notes or detecting themes in qualitative responses in open dialogues and structured questions.

Observations of the Child’s Emotional and Behavioral Functioning

Observations of the child in diverse environments, including the classroom, serve to evaluate behavior, attention, and social interactions. This assessment often involves the use of standardized measurement tools, such as questionnaires and rating scales, which are completed by parents, teachers, and/or observers. These tools, such as the Behavior Assessment System for Children (BASC), Conners Comprehensive Behavior Rating Scales (CBRS), Social Skills Improvement System (SSIS), Vineland Adaptive Behavior Scales (Vineland-II), and Child Behavior Checklist (CBCL), offer valuable insights into the child’s functioning across various domains. Although many rating scales already have automated scoring, generative AI and computer vision technologies could support practitioners with this process for those that do not. Further, given rigorous development, AI using computer vision could provide behavioral observations from within the classroom or at home and “flag” particular moments that may be of concern to practitioners to review.

Cognitive Assessment

Standardized cognitive tests are administered to assess the child’s intellectual functioning, problem-solving skills, memory, and other cognitive abilities. Standardized cognitive tests administered in psychological assessments vary in format and content. While some tests may be administered in a written format, others involve tasks that are verbal, visual, or hands-on. For example, tests like the Wechsler Intelligence Scale for Children (WISC) typically include a combination of verbal and non-verbal tasks, such as answering questions orally, solving puzzles, and arranging objects. Other assessments, like the Woodcock–Johnson Tests of Cognitive Abilities, may incorporate both written and hands-on tasks to evaluate problem-solving skills, memory, and other cognitive abilities. Additionally, advancements in technology have led to the development of computerized cognitive assessments, which may involve interactive tasks presented on a computer screen. Overall, standardized cognitive tests utilize a variety of formats to assess different aspects of intellectual functioning and cognitive abilities in individuals.

Academic Assessment

Academic achievement tests are administered to evaluate the child’s proficiency in essential areas such as reading, writing, mathematics, and other relevant subjects. These assessments come in various formats, ranging from standardized sets of questions to task-specific assignments. For instance, reading and mathematics skills may be assessed through automated tests featuring predetermined questions or tasks aimed at measuring comprehension and problem-solving abilities. In contrast, writing proficiency typically involves tasks such as essay composition, grammar exercises, or creative prompts. Moreover, discussions with teachers are frequently utilized to provide supplementary information, enriching the assessment process with insights from classroom performance and teacher observations.

Sensory and Motor Assessments (If Needed)

If there are concerns about sensory processing or motor skills, further assessments may be initiated to delve into these areas. These evaluations aim to gauge sensory processing abilities and motor coordination, pinpointing any potential challenges or areas requiring attention. Often, such assessments involve collaboration with speech-language therapists, occupational therapists, audiologists, or other specialists to gain a comprehensive understanding of the child’s needs and tailor interventions accordingly.

For all the aforementioned standardized testing, AI could provide scoring automation and preliminary reports, but psychologists must check for bias, errors, and validity of AI-generated outputs. However, there are current limitations to generative AI applications in assessment, as the requisite technologies have not been thoroughly investigated for applications in the context of psychological assessment. As such, the preliminary applications of generative AI would be summations of assessment in written format. For example, describing how a student performed in a particular academic domain relative to percentiles in plain English.

Data Analysis and Interpretation

Following the completion of assessments, the assessment team embarks on the crucial phase of data analysis and interpretation. This process involves meticulous examination and synthesis of the collected data to discern patterns, trends, and meaningful insights regarding the child’s strengths, weaknesses, and areas of concern. Through systematic scrutiny, the team identifies key findings indicative of specific learning or behavioral challenges, delineating the child’s cognitive, academic, and socio-emotional profile with precision. Various statistical analyses, qualitative assessments, and comparative evaluations may be employed to elucidate the significance of observed patterns and deviations from expected norms. Moreover, contextual factors such as environmental influences, developmental history, and individual differences are carefully considered to ensure a comprehensive understanding of the child’s unique circumstances. Through this rigorous analytical process, the assessment team elucidates the multifaceted nature of the child’s functioning, laying the groundwork for targeted interventions, tailored support strategies, and collaborative decision-making with parents and educators. As previously indicated, this is the primary support that generative AI can currently offer, to assist in streamline the scoring, and patterns across different assessments. Further, generative AI could create charts for parents, teachers, and students, and help explain percentiles and strengths and weaknesses in plain language. However, once again, practitioners are responsible for final interpretation, integrating context, and ensuring ethical oversight.

Feedback, Recommendations, and Intervention Planning

Following the assessment process, the findings are communicated to parents, often accompanied by a detailed report outlining recommendations for interventions, accommodations, or further assessments if necessary. Subsequently, an intervention plan is formulated based on the assessment results to address the child’s specific needs. This plan may include educational accommodations, therapeutic interventions, or specialized instruction tailored to support the child’s development and well-being. AI could provide intervention development feedback and insights. However, recommendations must always reflect human professional judgment and ethical considerations. Generative AI could be a useful tool, with correct training, to allow schools that lack the ability to implement a particular recommendation to more easily find alternatives, thus increasing the efficacy for some psycho-educational assessments. This aspect is particularly relevant for schools in rural Canada that may lack support compared to their urban counterparts.

Follow-Up

The process of assessment doesn’t conclude with the delivery of results; rather, it extends into the critical phase of follow-up. Recognizing the dynamic nature of a child’s development, periodic follow-up assessments may be recommended to track progress, monitor the effectiveness of interventions, and make adjustments to intervention plans as necessary. These follow-up assessments serve as checkpoints to gauge the child’s growth, identify emerging needs or areas of improvement, and refine intervention strategies accordingly. Moreover, they provide an opportunity for ongoing collaboration between parents, educators, and the assessment team, fostering a supportive and collaborative approach to addressing the child’s needs. Throughout the entire assessment process, the well-being of the child remains paramount, and every effort is made to conduct assessments in a supportive, respectful, and child-centered manner. By prioritizing the child’s welfare and ensuring a nurturing environment, the assessment process aims to yield accurate, meaningful results that contribute to the child’s holistic development and academic success.

AI systems could act as a responsive tool for parents, teachers and students to receive real-time monitoring and support. This could be done through the use of an online chat or through uploaded documents from parents and teachers that an AI could analyze based on the Psycho-Educational Assessment. This would allow AI to track trends or flag areas for review. However, such tools have not yet been validated in this specific context. Similar tools have been tested in clinical settings, demonstrating efficacy with minor variance (Beavers et al., 2023; Chen et al., 2025; Hasei et al., 2024).

AI Programs Supporting Educational Assessments

AI has become increasingly integral in enhancing the accessibility and accuracy of educational assessments. Several AI-powered tools are being used to streamline evaluation processes, provide personalized feedback, and improve learning outcomes. Below are key programs that specifically focus on educational assessments and their benefits. However, it is important to understand the context and development of these tools. They are not all equally situated within the scientific literature. Further, the following are what the tools can achieve in some contexts given the specific applications and situations. Whereas, sometimes these tools may be ineffective or even counter productive when their applications are beyond the developers intended scope. Thus, the following is simply to highlight some tools where AI is already being integrated. Lastly, when tools are not meeting their declared abilities it is important for practitioners, students and stakeholders to make changes aligned with the tools intended goals.

1. Gradescope: Currently utilized in various post-secondary institutions, Gradescope leverages AI to assist educators in grading student assignments. It improves grading efficiency by allowing educators to upload specific grading criteria, which the program then uses to assess student work. Additionally, Gradescope provides feedback and identifies trends in student performance, helping teachers to adapt instruction to student needs.

2. DreamBox learning: This adaptive learning platform uses AI to personalize math and reading instruction for students. DreamBox continuously assesses student progress and dynamically adjusts the curriculum to align with their learning pace and needs. By tailoring instruction in real-time, it maximizes student engagement and learning outcomes.

3. Pearson’s Q-interactive: This digital assessment tool utilizes AI for scoring and interpreting cognitive and academic assessments, such as the Wechsler Intelligence Scale for Children (WISC) and the Wechsler Individual Achievement Test (WIAT). The platform provides digital administration of tests, which significantly reduces the risk of human error during scoring, particularly in time-sensitive tasks. By automating these processes, Q-Interactive increases efficiency and allows educators and professionals to focus on interpreting results and making educational recommendations.

4. Cambridge cognition’s CANTAB suite: The CANTAB suite offers digital cognitive assessments, designed to eliminate variability and inconsistency in test administration. By collecting and analyzing performance data, the system identifies cognitive deficits and tracks changes over time, providing essential insights into student learning progress. This tool is particularly valuable in monitoring cognitive development and identifying areas where targeted interventions are necessary.

These programs highlight how AI technologies such as machine learning, natural language processing (NLP), and adaptive algorithms can impact educational assessments. By enhancing the precision, efficiency, and personalization of assessments, AI supports educators in offering more tailored instruction and improving student outcomes.

Sattler’s Pillars of Assessment: A Foundation for Generative AI Integration

Beyond the specific applications of generative AI within psycho-educational assessment, Sattler’s Four Pillars of Assessment serve as a fundamental framework widely applied in dealing with clients facing various psychological challenges. These four critical pillars, namely norm-referenced tests, interviews, behavioral observations, and informal assessments, offer a comprehensive understanding of an individual’s unique profile (Sattler, 2001). Building upon Sattler’s established framework, we now explore how the integration of generative Artificial Intelligence (AI) transforms each pillar, enhancing the assessment process while addressing the evolving landscape of educational psychology.

Norm-Referenced Tests and Generative AI

Norm-referenced tests play a crucial role in evaluating an individual’s performance in comparison to a standardized group. The integration of generative AI elevates these assessments by not only automating scoring but also revolutionizing the presentation of tasks. AI-driven adaptive interfaces can support engagement and tailor tasks to the child’s interests. However, it is important to note engagement is influenced by multiple factors, including motivation, test anxiety, and professional design oversight and these adjustments and the range of adjustments would have to be norm validated (Garcia et al., 2022; Lee & Wang, 2018).

Interviews Reimagined with Generative AI

Inspired by the innovative work of Shapira et al. (2021) in psychotherapy research, generative AI in interviews extends beyond structured data analysis (Martin & White, 2017). Virtual avatars or chatbots, crafted by AI algorithms, serve as interactive conversation partners, supporting interviews by assisting in note-taking or highlighting response patterns; human clinicians remain essential for building rapport and interpreting nuanced responses.

Ethical Considerations in Generative AI Integration

As generative AI integrates into educational assessments, the manuscript recognizes the paramount importance of addressing associated ethical considerations. Informed by Sattler’s emphasis on responsible assessment practices, the guidelines proposed in this manuscript delve into the need for a more in-depth exploration of potential ethical dilemmas, privacy concerns, and implications for vulnerable populations. The guidelines emphasize transparency and explainability in AI decision-making processes, ensuring the responsible and ethical use of AI in educational psychology assessments (Johnson & Davis, 2018).

Observations Transformed by Generative AI

Generative AI introduces a paradigm shift in the way observations are conducted (Chen et al., 2020). By creating immersive simulations or scenarios that replicate real-life situations, AI can augment observational data via simulations, though naturalistic observations remain essential for capturing authentic behavior. This innovative approach ensures a more engaging and meaningful assessment experience.

Informal Assessment Procedures Elevated with Gamification

Gamification takes center stage in the realm of informal assessments, facilitated by generative AI (Gomez et al., 2019; Miller & Brown, 2018). Educational games designed through AI algorithms not only capture essential data but also provide an enjoyable experience. This can enhance engagement, but assessment validity and reliability must be carefully maintained through ethical gamification and professional oversight. In navigating the integration of generative AI into educational psychology assessments, this manuscript intricately weaves technological advancements with established assessment principles, ensuring a harmonious evolution that prioritizes ethical considerations, engagement, and meaningful insights into the child’s unique profile. Personalized Learning Platforms: AI algorithms analyze individual student goals and preferences, paving the way for personalized learning platforms (Smith et al., 2020). These platforms can support personalized learning and highlight preliminary diagnostic information, which should then be interpreted and validated by professionals, transforming assessments into integrated and enjoyable components of the learning experience (Clark & Miller, 2019).

Challenges, Limitations, and Future Research

Despite potential benefits, the integration of generative AI in educational psychology assessments presents several challenges. These include the risk that AI could reinforce existing biases if datasets are not diverse or regularly audited, the necessity for continuous updates and validation of AI algorithms to maintain accuracy and fairness, and the potential for over-reliance on technology at the expense of professional judgment. Further, AI’s limitations in socio-emotional assessment should be recognized; outputs are supportive tools, not replacements for human judgment. It is also important to consider that generative AI and AI broadly currently has massive variance in efficacy, as some studies report massive success in assessment and human levels of variance, conversely other studies report that AI cannot perform in a standardized manner and has major variance, thus future research that evaluates efficacy for specific applications is paramount. Lastly, and perhaps the largest barrier for generative AI in any field, is access to data. Generative AI fundamentally requires vast amounts of data, thus the problem of data accessibility arises. This is of particular difficulty as psychological assessment data is considered confidential, thus two different opportunities arise. The first is to allow clinicians to ask clients to have their data used for training for generative AI, given the data is well anonymized. However, this could create a problem in terms of biasing particular AI agents; as such, this would create the need to ensure that data is being entered proportionally. A secondary option within the same vein would be to create particular agents that are specialized to particular diagnoses. The secondary major approach, and perhaps the more immediately useful for researchers, is to simulate the data. Increasingly, AI agents are being trained by simulated data created by other AI agents; thus, this could be used to simulate many profiles that meet a variety of different diagnoses and, in turn, how to report these different scores in terms of rating scales later. This has been done in a variety of disciplines to high levels of success (Sun et al., 2023; Yoon et al., 2023). However, its efficacy remains unknown for psychological assessments. Acknowledging these challenges is essential for responsible and informed implementation. Future research should consider additional involvement of practitioners to ensure that, in the case of simulated data or data labeling broadly, language variation is adequately accounted for in future models. Additionally, evaluation of the efficacy of generative AI in altering assessment materials or creating practice materials should be evaluated, as this would significantly benefit educators in supporting diverse student populations. Beyond assessment materials, future research should consider if generative AI can adjust course materials for a struggling student. For example, a student reading below grade level could receive a unique benefit if in class materials were adjusted to their grade level, assuming they still received the needed support and intervention, thus benefiting their social emotional development.

Proposed Guidelines and Ethical Framework

1. Human-centric approach: Generative AI should enrich engagement without overshadowing the diagnostician’s role emphasizing that engagement improvements rely on synergy with practitioner oversight and ethical design (Jones & Garcia, 2022). Prioritizing the child’s experience ensures a collaborative synergy between technology and human expertise.

2. Ethical gamification: Gamification is employed ethically, ensuring that engagement complements, not compromises, assessment validity (Miller & Brown, 2018). A delicate balance between engagement and accuracy is crucial for the responsible use of gamified assessments.

3. Transparency and explainability: AI algorithms used in assessments are designed to be transparent and explainable, empowering diagnosticians to comprehend and validate generated insights, fostering trust in the assessment process recognizing that some complex AI models may have limited explainability (Wang & Chen, 2019).

4. Continuous collaboration: Ongoing collaboration between educational psychologists and computer scientists is imperative for refining AI algorithms (Martin & Gomez, 2020). This collaboration ensures that AI aligns with the evolving understanding of human behavior, fostering a dynamic and adaptive assessment landscape.

5. Privacy and security: Robust measures are in place to safeguard the privacy and security of sensitive data (Yang & Lee, 2017). Stringent compliance with ethical standards and regulations ensures responsible and ethical use of AI in educational psychology assessments (Johnson & Davis, 2018).

6. Training and professional development: Joint training programs for diagnosticians and computer scientists promote a shared understanding of the assessment process (White & Miller, 2019). This collaborative approach ensures effective and ethical integration of AI, harnessing the strengths of both fields.

7. Clarify jurisdictional differences: Compliance with privacy and education laws varies by region (e.g., FERPA in the U.S., GDPR in the EU, PIPEDA/FIPPA in Canada).

8. Human responsibility: AI outputs inform decisions, but ultimate accountability lies with practitioners, developers, and institutions; redress mechanisms (e.g., appeals, review boards) should be established.

Relevance to the Practice of School Psychology

Generative AI could not only streamline the assessment process but also holds the potential to revolutionize engagement for children. The proposed guidelines and ethical considerations outlined in this manuscript, informed by the seismic shift highlighted by Duff and Roberts (2016) and inspired by insights from psychotherapy research (Shapira et al., 2021), provide a roadmap for educational psychologists and AI developers to collaborate responsibly. By maintaining practitioner oversight, ensuring transparency, auditing for bias, and prioritizing the child’s experience, AI-supported assessments can be both efficient and meaningful. This approach fosters a human-centric, ethical, and accountable integration of generative AI into educational psychology assessments, enhancing diagnostic precision while safeguarding the developmental, emotional, and educational needs of the child.