Why ChETL is designed the way it is
ChETL starts from one conviction: engineering competence is the ability to reason and to judge, to choose assumptions, weigh models and interpret results, not merely to compute. This page explains how that conviction shapes the platform, and separates what exists today from what remains an aim.
The gap between an answer and an understanding
Chemical engineering students already have good resources, and each solves part of the learning problem well. Used on its own, each tends to leave a gap compared with an excellent classroom, where a teacher explains a method, works a problem at the board, reads the answer back into engineering meaning and then questions the class on it.
| Resource | What it provides | A common limitation when used alone |
|---|---|---|
| Textbooks | rigorous theory and validated models | a printed text is fixed: it cannot respond to a learner’s question or recompute a case under new conditions |
| Recorded lectures | expert explanation at scale | watching can feel like understanding without the learner having to act; that depends on how the material is used |
| Process simulators | industrial-grade computation | they are built for practice rather than teaching, so substitutions and intermediate steps are often not presented to the user |
| Computational notebooks | interactive, executable exploration | most assume some programming ability, which can be a barrier for beginners |
| General AI assistants | instant conversational explanation | unless the learner supplies it, a general assistant does not know which model, assumptions and numbers the learner is working with, and its answers may be wrong |
The risk is clearest with general-purpose AI: it is good at producing answers, while engineering education is about developing reasoning. A student can obtain a plausible, or even correct, number without understanding the steps that produced it. ChETL exists to close that distance.
Reasoning and judgement as the aim
A competent engineer does more than evaluate a formula. They decide which model applies, which assumptions are safe, whether a result is plausible and what it implies for a decision. Those skills are built by making such decisions, not by watching them made.
ChETL therefore takes the structure of a good lesson as its blueprint: purpose first, then a narrative that builds the method, a worked demonstration, the freedom to try other cases, a reading of what the result means, a chance to ask questions, and a check of understanding. At each stage the learner is asked to reason, not only to observe.
One calculation, three explanations
Each topic separates three questions about the same calculation, and gives each its own surface so that none is diluted by the others:
- Theory: why and how
- A written chapter that builds the general method, including where the governing equations come from, and maps the solution strategy before any number appears.
- Blackboard: the worked arithmetic
- The engine’s numerical path for the case the learner configured: knowns with their sources, assumptions, then each step as formula, substitution and result, with units. It contains computed numbers only; symbolic derivations belong to the Theory.
- Interpretation: what it means
- The result read back into physical and engineering meaning for this case, closing with the model’s validity and its limitations.
The philosophy describes a board demonstration in four acts: assumptions, derivation, computation and interpretation. In the platform those acts are deliberately divided. Assumptions and computation are on the Blackboard, derivation is in the Theory, and meaning is in the Interpretation. A fact appears in the form that suits it: an equation with numbers on the Blackboard, a general method in the Theory, a plain-language conclusion in the Interpretation.
Plot and Data show the same calculation across a range of conditions, and the How it works tour shows all the surfaces for one real case.
Transparent calculation, one numerical authority
Every calculated result and intermediate value that the platform presents as a ChETL result comes from one calculation engine. The Blackboard, the plot, the interpretation and the data table read the same result; none of them calculates on its own. That is what allows a learner to trace a plotted point back to its Blackboard step and trust that both describe the same calculation.
The guarantee has limits. Numbers written in AI Tutor replies or in generated assessment questions are not engine results, and quoted reference data, textbook values and illustrative figures on this website are separate from the calculation.
Transparency includes honesty about what cannot be calculated. If a model has no parameters for a component, or a state lies outside what a route can represent, the platform says so and explains why. It does not quietly substitute an ideal model, invent a value or draw a point it did not compute.
Exploration and trends
A lecture rarely has time for more than one or two worked examples. On ChETL a learner can change one input, a temperature, a composition or a model, and see every dependent value change with it. The question then shifts from “what is the answer?” to “why does the answer change?”, which is where engineering thinking begins.
A trend is still a model result for a stated range. The Interpretation tab describes the trend for the conditions calculated and does not present it as a general rule.
Assessment as part of learning
Assessment is meant to reveal and correct a misconception before it becomes permanent, not only to produce a score. In the platform today it offers:
- three levels: Remembering (definitions, symbols, equations), Comprehension (meaning, trends, why a behaviour occurs) and Application (choosing a model or route, reasoning from a supplied value);
- two modes: Interactive Learning, which marks each answer immediately with an explanation, and Classroom Simulation, an independent attempt with feedback only after the set is complete.
Questions are tied to the active topic and, where relevant, the solved case. They may be generated by AI and are screened by automated format and numerical checks; a topic-specific curated set is used when generation fails. Generated questions are not described as human-reviewed. The wider ladder of learning, through analysis and evaluation to design, guides how the Theory is written; it is not claimed for the current assessment, and ChETL does not claim measured learning gains.
The role of AI
AI is not the teacher, the textbook or the calculation engine. It is an assistant working inside a designed lesson, and its boundaries are deliberate:
- Knowledge stays grounded in sources. Theory and models rest on cited textbooks, literature and reference data. Generative AI is not treated as a scientific source.
- Official results stay with the engine. The AI Tutor explains the case that is open; nothing it writes changes the calculated results, and it is instructed to treat them as fixed and to send requests for other conditions back to the configuration. Its replies are still generated text, so a number written in a reply can be wrong and should be checked against the calculated results.
- Context is supplied, not guessed. The Tutor receives the topic, model, conditions and results of the current case, so the learner does not need to describe them.
Tutor answers are generated dynamically, can vary and may be inaccurate; the platform labels them as such and points back to the Theory, the calculated results and cited references. The Tutor is available to signed-in users within a daily question allowance. AI tools may also assist drafting and consistency checks of written content under human editorial responsibility; such checks are quality assurance, not peer review.
Human responsibility
Scientific and editorial responsibility for ChETL’s content rests with the ChETL Scientific Editor, the project owner. Development and review records are kept under version control, and material errors are corrected when found; the route is on Contact. Responsibility for an engineering decision stays with the engineer: ChETL’s results are model estimates for learning, not design values. More detail is in Sources, review and limits.
Design principles
- Transparency
- reasoning is visible; each Blackboard value is shown as formula, substitution and result, with units.
- Rigour
- real equations, stated assumptions and models used within their ranges, computed by one engine.
- Continuity
- each topic builds on earlier ones and prepares for later ones; a concept is derived once and referred to elsewhere.
- Clarity
- restraint over decoration, so the path from purpose to understanding stays legible.
- Accessibility
- plain English, so that learners whose first language is not English can follow demanding ideas without fighting the prose, and no programming prerequisite.
- Respect
- for the discipline, for teachers whose craft the platform draws on, and for the intelligence of the reader.
Present practice and long-term aims
A philosophy can describe more than a platform yet does. The two are kept apart here.
In the platform today
- One course, Chemical Engineering Thermodynamics, with five available modules; Thermodynamic Cycles is upcoming.
- Each topic in one workspace: Theory, Blackboard, Plot, Interpretation, AI Tutor, Assessment and Data.
- Assessment at three levels in two modes, with no gradebook or learning-management-system integration.
- A context-aware AI Tutor with a daily allowance for signed-in users.
- One complete, independently reproduced public worked example so far.
Long-term aims, not current claims
- A learning environment spanning the core chemical engineering curriculum, such as transport phenomena, separations, reaction engineering, process control and design, on the same principles.
- Learning that continues at each person’s pace, well beyond what one classroom allows.
- The full progression of learning, from understanding to evaluation and design, supported throughout.
These are ambitions without dates. The near-term commitment is depth: each topic done well before the next is added.