Prof. Sandeep Goyal, Professor, Information Systems and Business Analytics, LM Thapar School of Management, Thapar University, Patiala
A student is the most difficult system I have ever tried to debug. There is no documentation. There are no error logs. And the feedback loop runs on a delay of years, so I rarely learn whether a fix worked until long after the student has left the room. I spent two decades building and repairing large software systems before I taught my first class, and nothing in that work prepared me for how much harder a young mind is to read.
It also taught me the one thing that carries over cleanly. In any complex system, the visible output is never the whole story. The lecture I deliver is an interface. The learning that matters happens in the system underneath, and that system is a person, not a syllabus.
For most of its history, teaching is optimized for one function: closing an information gap. The professor held the knowledge, the student lacked it, and the lecture transferred the same. That design made sense when knowledge was scarce and expensive to reach. That design is now obsolete. A student pulls a cleaner explanation of almost any concept from a phone in the time it takes me to open my slides. If information is the only thing I compete on, the machine wins—every time.
The signal is the relationship, not the content
An analyst learns early to separate signal from noise. Spend enough years staring at what actually drives an outcome, and you stop trusting the loud, obvious variable. Content is the loud variable in education: easy to measure, easy to deliver, and far less decisive than we assume.
One of the largest surveys of its kind put real numbers on this. Gallup and Purdue (2014) asked more than 75,000 college graduates which parts of an education actually predicted a good working life afterward. The strongest signal was not the ranking of the school. It was not the grades on the transcript. It was whether a graduate strongly agreed that a professor cared about them as a person.
The effect held up when they measured it. Graduates who felt genuinely cared for were nearly twice as likely to be engaged in their work years later, and much more likely to be
thriving in their wellbeing. Later work sharpened the finding: graduates who had a mentor who pushed them toward their goals were twice as likely to be engaged at work, and the mentor was most often a professor (Strada and Gallup, Mentoring College Students to Success, 2018). One human relationship, measured years downstream, still moved the outcome variable that every ranking pretends to predict.
The gap between what matters and what we deliver
Here the same study delivers its gut-punch. Only 27% of graduates strongly agreed they ever had a professor who cared about them as a person (Gallup and Purdue, 2014). The 2018 follow-up found the gap just as wide: fewer than half of graduates said they had any mentor at all during college (Strada and Gallup, 2018). Yet this experience is available to only a minority of the students. We have become very good at delivering content. We have been less consistent at building relationships with the students receiving it.
India has at least named the redesign, and recently started to fund it. The National Education Policy 2020 recasts teachers as mentors and facilitators rather than instructors, and in March 2024 the government issued the National Mission for Mentoring Bluebook to turn that language into a working framework (Government of India, NEP 2020; NCTE, National Mission for Mentoring Bluebook, 2024). Policy sets the target. Closing the gap is the work, and it happens one relationship at a time.
Mentoring is debugging without the logs
So what changes in practice? The work moves off the syllabus and into the parts of a student a lecture never touches.
The first task is to listen for the signal a student does not send. A quiet student in the third row may be bored, or may be coming apart under a pressure no one in the room knows about. The output looks identical from the front. Only attention, paid deliberately, tells the two states apart.
The second is to coach the process, not patch the output. When a student hands over a broken analysis, marking the error is the quick fix, and it teaches nothing durable. The better move is the question that lets them locate the fault themselves, because the capability worth building is not this one corrected answer. What lasts is the habit of interrogating their own reasoning long after the professor is gone.
The third is to treat the whole person as the system, not the graded fraction of them. The student in the front row is also an individual under family pressure, a first-generation graduate carrying a village's expectations, a capable young person quietly unsure they belong in the room. Ignore those inputs and you are tuning a transcript while the actual system struggles somewhere no one thought to look.
The lesson the systems taught me
Two decades of running project delivery teams left me with one conviction that survives the move from industry to campus. People do not deliver their best because you specified the task well. They deliver it because they believe you are genuinely invested in their growth, and they refuse to let that trust down.
A student runs on the same logic. The moment they sense that I care about who they are becoming, and not just what they score, the whole dynamic re-architects itself. Attendance stops being a rule to enforce. Effort stops being a transaction. They begin bringing me the questions that actually matter, the ones about direction and doubt and fear, because they trust I will treat those with the same seriousness as any exam.
What the machine cannot do?
It is worth being precise about where the machine wins, because the honest list is long. A model explains a concept more patiently than I can, at any hour, in any language, and is never tired of the fifth version of the same question. It drafts, it summarises, it tutors, it grades a multiple-choice set in the time I take to find my pen. On the pure transfer of information, the task teaching was built around, I have already lost, and I am not interested in pretending otherwise.
So, I have had to ask a harder question: once the machine does the explaining, what is the professor actually for? Whatever it is, it lives past explanation, in everything the model has no access to. A model holds no stake in this particular student. It cannot want something for her that she does not yet want for herself. It does not notice that the sharp student in the third row has gone quiet for three weeks, because it was never in the room. It carries no memory of having been wrong, no scar tissue, nothing at risk when it advises. It answers the question asked. It cannot decide that the question asked is the wrong one and that the student is avoiding the real one out of fear.
That is the ground we do not surrender. Judgment under real stakes, the courage to tell a capable student a hard truth about one’s work, the read on a person that comes only from caring what becomes of them. The machine optimises for the answer. A teacher is accountable for the human. The more fluent the model becomes at the first, the more clearly the second stands out as the whole of the job.
Where I land?
The classroom ahead will not belong to whoever owns the best slides, or the best model. Information is free now, infinite, one query away, and explanation is close behind it. What stays scarce, and always will, is a teacher who reads a student clearly and refuses to let them ship a smaller version of themselves.
That is the teaching that happens beyond teaching. Across every system I have worked on, in code and in people, one rule has held. Users forget the features you shipped. They never forget how the thing made them feel while they used it. Students are no different. They may forget much of what we teach. What they are less likely to forget is how a teacher made them feel about themselves and their possibilities.
E-305, Industrial Area, Sahibzada Ajit Singh Nagar, Punjab 160055, IN
-500-×-500-px.jpg)
