[KongchangAI]
· 2 min read· 1,060 words

Caltech Professor Stops Grading Homework Due to AI Prevalence

Caltech Professor Stops Grading Homework Due to AI Prevalence

A Caltech professor stopped grading homework because students widely use AI to complete assignments, sparking discussion on reforming university assessment systems.

A Caltech professor decided to stop grading homework because nearly all students are using AI tools to complete assignments, eroding the effectiveness of homework as an assessment method. Homework, completed outside class and unsupervised, is the assessment component most vulnerable to large language models. Abandoning grading means assessment pressure shifts to exams, projects, and other formats, but grading itself also serves a feedback function; stopping grading may sacrifice feedback quality. This event reflects that university assessment system adjustments lag behind the pace of AI tool adoption, forcing course design to once again answer how to prove that students have learned.

A Caltech professor has decided to stop grading homework, pointing directly at the reality of today's classroom: nearly all students are using AI tools to complete assignments. The news sparked discussion on Hacker News and once again brought the tension between university assessment systems and technological tools to the forefront.

From Grading Homework to Abandoning It: What Happened

According to discussion on Hacker News, this Caltech professor chose to stop grading homework, with the core motivation being that students widely use AI to complete assignments, eroding the effectiveness of traditional homework as an assessment method. The information itself is limited—no course name, professor's name, or specific alternative plan was disclosed—but the issue it touches is concrete enough: when AI can easily generate seemingly acceptable homework answers, how much meaning does grading each one still have?

This is not an isolated complaint but a signal. As part of the teaching process, homework has long served three functions: consolidating knowledge, exposing comprehension gaps, and providing a basis for grading. When two of these functions are significantly weakened by AI tools, a professor's rational response is either to change the format of homework or to abandon incorporating it into grading.

Why Homework Is the First to Fall

The difference between homework and exams precisely determines why it is most vulnerable to AI disruption. Exams typically have proctoring, time limits, and a closed environment; homework is completed outside class, where students can freely use any tools, including large language models.

A professor cannot determine whether a homework submission reflects a student's thinking or a model's output, especially when the answer itself is correct and the writing is fluent. What the grader faces is no longer evidence of comprehension level but a sample of generation quality. In this situation, continuing to grade homework is equivalent to basing scores on unverifiable signals.

Interestingly, the problem is not whether students are "cheating." Even if students use AI to help understand problems and organize their thinking, homework is no longer a product completed independently. The default premise of the assessment system—that homework reflects students' own abilities—no longer holds without additional verification methods.

A large language model (LLM) is a type of deep learning model trained on massive amounts of text, capable of generating coherent natural language based on context. Its core capabilities include understanding instructions, reasoning, and creation, with typical examples such as the GPT series. LLMs generate text by predicting the probability distribution of the next word; after large-scale pre-training and fine-tuning with reinforcement learning from human feedback (RLHF), they can perform tasks such as question answering, writing, and code generation. In educational settings, LLMs can quickly generate clear, well-structured, and fluent answers that are difficult for traditional plagiarism detection tools to identify, directly undermining the reliability of extracurricular homework as proof of independent ability.

Chain Effects on the Teaching Assessment System

This one decision by the professor points to a larger structural adjustment. If homework exits grading, the assessment pressure of the course will shift to exams, projects, classroom participation, and other components. Whether these components can bear the load depends on whether they are equally easy to bypass with AI tools.

From the atmosphere of the Hacker News discussion, the community's attention focused on two directions: first, whether universities will accelerate their shift toward oral exams, live coding, formative assessment, and other formats that are harder for AI to replace; second, whether abandoning homework grading amounts to abandoning teaching feedback. The latter concern is not unfounded—the value of homework is not only grading; the grading process itself is also a channel for students to receive targeted feedback. A professor stopping grading may simultaneously sacrifice feedback quality.

Currently, public information is insufficient to determine whether this professor has designed an alternative assessment plan. But the decision itself shows that, at least in this course, the assessment function of traditional homework has been judged ineffective.

Formative assessment and summative assessment are two basic types of educational evaluation. The former focuses on performance during the learning process, such as classroom participation, staged assignments, and project progress, aiming to provide feedback to improve learning; the latter focuses on outcomes at the end of learning, such as final exams and final projects, used for grading or certification. Homework usually serves both functions. When AI tools render the summative assessment function of homework ineffective, educators may turn to more formative assessment—such as oral defenses, live coding, and learning logs—because these formats are harder for AI to replace and can capture students' thinking processes. However, formative assessment places higher demands on teacher time and course scale, and its promotion faces practical constraints.

Tool Prevalence Forces Assessment Reform

The speed at which AI enters the classroom is faster than the adjustment speed of assessment systems—this is the common situation facing universities today. A single professor's response may not be the standard answer, but it makes a previously avoidable question concrete: when students widely use AI, course design must once again answer "how to prove that students have learned."

For teachers, feasible directions include changing homework to be completed in class, adding oral defense components, and replacing one-time submitted assignments with projects that require process records. These solutions each have costs and cannot completely eliminate AI's influence. For schools, the more fundamental question is how to redefine the credibility of academic assessment in an environment where tools are widespread.

This one decision by the Caltech professor, in the short term, only affects the grading method of one course; in the long term, it is a public statement that the assessment system must face.

One-line Summary

A Caltech professor stopped grading homework because students widely use AI to complete assignments, sparking discussion on reforming university assessment systems.

Paragraph Summary

A Caltech professor decided to stop grading homework because nearly all students are using AI tools to complete assignments, eroding the effectiveness of homework as an assessment method. Homework, completed outside class and unsupervised, is the assessment component most vulnerable to large language models. Abandoning grading means assessment pressure shifts to exams, projects, and other formats, but grading itself also serves a feedback function; stopping grading may sacrifice feedback quality. This event reflects that university assessment system adjustments lag behind the pace of AI tool adoption, forcing course design to once again answer how to prove that students have learned.

#Caltech professor#AI homework#university assessment#large language models#educational evaluation reform#formative assessment#academic integrity
Share:

Related articles