Beyond Grading: Towards Intelligent, Ethical, and Learning-Focused Assessment

Reading Time: 9 minutes

For decades, the primary purpose of learning assessment was considered to be measuring, i.e., administering a test that verified whether the student understood and mastered the subject, and, many times, this defined the academic destiny of the students. 

However, today, this way of understanding evaluation is in crisis. In the 21st century, evaluation encompasses not only measurement but also accompaniment, understanding, and providing feedback on the learning process. In particular, advances in artificial intelligence (AI), adaptive learning, and data analytics have created new opportunities to design assessments that are fairer, more ethical, and personalized. While this is a positive development, it also raises dilemmas about the role of teachers, privacy, and student autonomy. 

From measurement to accompaniment

In his article From Tests to Current Evaluative Research (2003), Tomás Escudero describes the evolution of educational evaluation throughout the twentieth century, illustrating how it transitioned from a technical and quantitative process to one that is more complex, ethical, and pedagogical.

The author identified four primary stages that synthesize the paradigm shifts:

1. The measurement stage (early twentieth century):
   • Dominated by the psychometric paradigm, evaluation was conceived as a process of objectively measuring skills and performance. The model of the natural sciences inspired it; statistical precision and standardization were prioritized. The student was considered an “object of measurement,” and the purpose was to compare, classify, and predict performance. Figures such as psychologists Thorndike and Terman dominated this era, focusing on intelligence tests and normative tests.
2. The descriptive stage (mid-twentieth century):
   • Escudero’s analysis identified that during the second stage, with the expansion of mass education and the influence of Ralph W. Tyler, evaluation began to be described in terms of educational objectives. Tyler designed a curriculum development model that introduced the idea of assessment to check the degree of achievement of previously defined goals. This gave rise to planning by objectives and curricular evaluation. Still, the approach remained functionalist: products were evaluated, not processes.
3. The value judgment stage (1960s and 1970s):
   • Escudero points out that during the third stage, the contributions of Scriven and Stake led to a conceptual shift regarding evaluation: it ceased to be limited to measuring and began to consider well-founded value judgments about the quality and relevance of programs. Scriven introduced the distinction between formative and summative assessment, while Stake developed the responsive approach and promoted the use of case studies. The evaluator was recognized as much as an interpreter as a technician, and the idea that all assessment carries an ethical and political dimension was consolidated.
4. The stage of evaluative research (end of the twentieth century):
   • Finally, Escudero describes the consolidation of a holistic, interpretative, and participatory vision. Evaluation became integrated into educational research as a practice of understanding and transformation, where context, actors’ voices, and feedback take center stage. It is no longer a matter of qualifying results, but of accompanying learning processes and institutional improvement.

Escudero’s survey of assessment arrived at a key conclusion: evaluation is a social phenomenon, not merely a technical one. The author emphasizes that reducing it to only measurement impoverishes its meaning, because evaluating implies understanding, dialoguing, and making informed decisions to transform teaching.

For this reason, the author proposes conceiving the teacher as a “professional evaluator,” not just someone who applies tests, but also who mediates between information, interpretation, and pedagogical action. Evaluation, in this perspective, becomes a reflective, ethical, and collaborative practice that aims for continuous improvement. From this arises the need for the teacher to be not only a disciplinary expert, but also an evaluation professional, capable of observing, providing feedback, and accompanying.

Along these lines, Sánchez Mendiola expanded the proposal by introducing a model that assesses the “from,” “for,” and “how” of learning, which incorporates Escudero’s ideas into the contemporary context of higher education. “From” learning means certifying achievements and accrediting results; “for” learning guides teaching and the provision of feedback, and “how” learning refers to student self-regulation and reflection. These three approaches position the student at the center of the process, fostering a culture of continuous learning rather than merely emphasizing accountability.

Additionally, the author proposed that strategies and instruments should be aligned with observable and measurable learning outcomes and suggested replacing rote exams with tasks that simulate real contexts. Moreover, he believes that AI and learning analytics can enhance feedback when used for training purposes. 

The Power of Feedback and Interaction

On the other hand, authors John Hattie and Helen Timperley conducted a meta-analysis of more than 500 studies and 7,000 effects on variables that influence learning. In their synthesis, they found that feedback is one of the most potent factors for improving academic performance, with an average effect size of 0.79. This means that well-designed feedback can have almost twice the impact of teaching (average effect size 0.40). However, they advised that not all feedback generates improvement: its effectiveness depends on its nature, timing, and orientation towards learning.

The authors proposed a three-level feedback model that guides towards effectiveness:

1. Task level: focuses on correcting specific errors or the quality of the final product.
2. Process level: guides the student on how to improve their understanding or strategy.
3. Self-regulation level: promotes autonomy, metacognition, and self-evaluation.

Their research indicated that the second and third levels are the most powerful, as they encourage reflection and the transfer of learning, rather than simply superficial corrections.

In addition, their research found that feedback answers three key questions:

• Where am I going? (Feed up): clarifies goals and criteria.
• How am I doing? (Feedback): provides information on current progress.
• What’s next? (Feed forward): guides the following steps to improve.

This feedback cycle considers the past (feed-up), present (feedback), and future (feed-forward) aspects of learning, transforming assessment into a continuous process of orientation and growth.

The sociocultural view of Esterhazy and Damşa (2017) contributes to this cognitive perspective, as they rethink feedback as a practice of jointly constructing meaning rather than a one-way message. Their research with university students shows that the value of feedback arises from interactions: the student dialogues about the comments, interprets them in collaboration with peers or teachers, and incorporates them into their future work.

From this perspective, feedback is not an isolated event, but a learning path—a meaning-making trajectory—where interpretations, emotions, and expectations are negotiated. Learning occurs precisely in this intersubjective space, where error is no longer seen as a failure; instead, it is transformed into an opportunity for growth and development.

For Esterhazy and Damşa, feedback does not mean correction, but accompaniment. The teacher considers the evidence of performance and the possibilities for improvement, guiding the students to build their knowledge. Evaluation ceases to be an act of grading and becomes a dialogue, reflection, and co-authorship of learning.

Evaluative Literacy

Today, it is no longer sufficient to know how to apply instruments or record grades; it is a matter of understanding the pedagogical, ethical, and formative meaning of evaluation. Transforming assessment into a learning process requires teachers to develop a new professional competency: evaluative literacy.

Pastore and Andrade (2019) describe evaluative literacy as the integration of knowledge, skills, and ethical dispositions that enable teachers to use assessment in a fair, valid, and improvement-oriented manner.

Their three-dimensional model proposes interdependent aspects: cognitive or knowledge, practice or application, and ethics or values. The first involves understanding the theoretical, methodological, and psychometric foundations of assessment. This includes designing instruments aligned with learning objectives, recognizing different types of evidence, and selecting appropriate techniques for each context.

The practice or application part refers to the ability to use evaluation as a pedagogical tool, not just an administrative one. A teacher with evaluative literacy interprets the results, adapts their teaching, and provides continuous feedback. Assessment becomes a two-way learning process, where the teacher also learns from their students.

Finally, the ethical or values dimension is the most profound and often the least visible. Here, evaluating responsibly implies recognizing the power it has to affect students’ trajectories, emotions, and perceptions of competency. Evaluative ethics translates into transparency, inclusion, and contextual sensitivity. The teacher must reflect on why they evaluate and who benefits from their evaluation. These questions are just as important as deciding which instrument to use.

These dimensions seek to transform the technical vision of teachers as an applier of tests, redefining them as reflective agents who design, interpret, and act on evaluative information. In the words of Pastore and Andrade (2019), evaluative literacy is “situated knowledge”: it changes according to the cultural and institutional context, so it must be built collectively within academic communities.

Marzano and Kendall’s New Taxonomy

Katherine Gallardo (2009) expands on the conception of Pastore and Andrade by linking it to the New Taxonomy of Marzano and Kendall, a theory of human thought that surpasses Bloom’s static hierarchy. Instead of ordering processes by difficulty, Marzano and Kendall distinguish three mental systems that intervene in learning:

1. Cognitive system: responsible for processing information and solving tasks through different levels of processing: retrieval, comprehension, analysis, and use of knowledge.
2. Metacognitive system: sets goals, monitors progress, and regulates cognitive strategies.
3. Internal or self-regulatory system: connects the motivation, beliefs, and emotions that drive or inhibit learning.

In this framework, evaluation implies observing how the student thinks, self-regulates, and becomes emotionally invested in their learning. Thus, evaluation is not limited to the final product, but analyzes the processes that made it possible. For example, a well-designed rubric can assess not only the accuracy of an answer, but also the clarity of the strategy used or persistence in the face of difficulty.

Gallardo (2009) proposes that teachers articulate evaluation using the three systems: cognitive, to measure comprehension and performance; metacognitive, to encourage reflection on one’s learning; and internal or affective, to strengthen motivation and the sense of achievement. In this way, assessment becomes a comprehensive and humanistic practice, which addresses the what, how, and why of learning. 

When a teacher masters evaluative literacy and applies the New Taxonomy, their feedback goes beyond correcting errors and becomes a conscious accompaniment to the development of thought.

Artificial intelligence and adaptive learning

Artificial intelligence (AI) applied to education represents an unprecedented opportunity to personalize teaching and assessment. Its potential lies in the ability to analyze large volumes of data in real-time, whether it be results, interactions, response times, or error patterns, to adjust instruction to each student’s profile.

Models such as the Personalised Adaptive Learning and Assessment System (PALAS), developed by Palanisamy, Thilarajah, and Chen, illustrate this pedagogical revolution. This model integrates algorithms capable of diagnosing prior knowledge and comprehension gaps, generating personalized learning sequences, offering immediate and contextualized feedback, and updating learning paths based on individual progress.

Instead of following a linear, homogeneous curriculum, the student undertakes “personalized learning paths,” where each activity, question, or assessment dynamically adjusts to their performance. This makes evaluation a continuous process of automated measurement and feedback, rather than an end event, allowing for difficulties to be detected before they consolidate and for earlier interventions to be implemented.

Beyond technical efficiency, the actual value of these systems lies in their potential for equity. PALAS, for example, was designed as part of Singapore’s national AI strategy to reduce the education gap associated with socioeconomic background or special needs.

By offering personalized support, students who are progressing more slowly receive immediate reinforcement, while those who are progressing faster can access additional challenges. This transforms the ideal of “inclusive education” into a concrete, data-driven practice.

As the article The Ethics of Learning Analytics in Australian Higher Education points out, learning analytics necessitate explicit governance principles that regulate the collection, storage, and use of student information, ensuring transparency, informed consent, and institutional accountability. 

Although artificial intelligence and learning analytics are powerful tools, the real challenge is not in how the algorithms work, but in how people use them and for what educational purpose. AI can indeed strengthen formative assessment if it is used for providing feedback, rather than for monitoring. It can promote equity if algorithms are designed to detect and correct biases, not to reproduce them, in addition to freeing up teaching time if institutions prioritize pedagogical support over the simple generation of metrics.

Innovative education is not about delegating judgment to an algorithm, but about using data to make fairer, more nuanced, and context-sensitive decisions. In this sense, AI does not replace the teacher; instead, it amplifies their capacity.

When combined with evaluative literacy frameworks, learning analytics and adaptive systems become powerful tools for teaching, feedback, and learning supported by empathy, evidence, and purpose.

Ethics and Humanism in the Age of Learning Analytics

The report The Ethics of Learning Analytics in Australian Higher Education warns that while data can provide valuable insights into personalized teaching, it can also lead to surveillance risks, algorithmic discrimination, or a loss of student autonomy if there are no explicit governance frameworks in place.

The fundamental ethical question is not whether we should use data, but rather how and for what purpose we use it. Control-oriented analytics erode trust and transform education into a monitoring system; on the other hand, an accompaniment-oriented analysis enhances reflection, improvement, and inclusion.

Universities have the responsibility to establish institutional policies that govern the use of educational data, which include a clear delineation of responsibilities among teachers, designers, and analysts; periodic auditing of algorithms; interdisciplinary ethical review committees; and mechanisms for student participation in decision-making about the use of their information.

Likewise, algorithms must be explainable and interpretable, so that users understand how recommendations or predictions are generated. Analytics cannot replace teacher judgment, because understanding the context, emotions, and intentions of learning remains an irreplaceably human endeavor.

The ethics of digital assessment boil down to a crucial question: Who is educational technology serving? If the answer is “It is in the service of learning,” then AI can become a powerful ally in strengthening fairness, inclusion, and personalization.

The challenge is to design an ecosystem where data informs, but does not dominate; where technology assists, but does not control; and where every educational decision continues to be guided by empathy, trust, and a commitment to training.

Towards authentic and transformative assessment

In his most recent proposal, Sánchez Mendiola (2022) defines authentic assessment as one that reproduces the challenges of the real world and requires the student to transfer their knowledge to complex, interdisciplinary, and socially meaningful situations. For the author, the purpose is for assessment to stop being a filter and become a learning experience in its own right.

Unlike standardized tests, which focus on memorization or immediate results, authentic assessment values process, reasoning, and the contextual application of knowledge. For this, it is recommended to use strategies such as integrative projects, case studies, clinical or business simulations, digital portfolios, and innovation challenges. These methodologies foster skills such as critical thinking, collaboration, communication, self-regulation, and creativity, which constitute the core of 21st-century skills.

Artificial intelligence can enhance this form of evaluation if it is used pedagogically. For example, adaptive tools enable the analysis of performance patterns and provide immediate feedback. Learning analytics platforms help visualize both individual and collective progress, while digital simulation environments facilitate authentic experiences in safe and controlled contexts.

However, the real transformative power stems not from technology, but from how experiences are designed and data is interpreted. An authentic evaluation requires the alignment of curricular objectives, learning tasks, and performance criteria, as well as the teacher’s ethical commitment to accompany the processes with meaningful feedback.

More than measuring results, authentic assessment seeks to understand learning as an integral phenomenon, encompassing emotion, motivation, and a sense of achievement. Evaluation in the twenty-first century moves away from the paradigm of measurement and is oriented towards accompaniment, personalization, and ethical reflection.

Its purpose is not to classify, but to understand and enhance human development. Integrating artificial intelligence, learning analytics, and cognitive frameworks, such as the New Taxonomy, does not imply dehumanizing education; instead, it reaffirms its essence: teaching with empathy and evaluating with purpose.

Translation by Daniel Wetta

This article from Observatory of the Institute for the Future of Education may be shared under the terms of the license CC BY-NC-SA 4.0