The Work/Should AI Grade Your Work?

Should AI Grade Your Work?

Human grading isn't neutral either. Neither answer is the easy one.

AI in BusinessResearch
Overview
Proposed AI governance model

Should universities let AI grade student work? The honest answer isn't "yes" or "no" — it's about who's actually accountable when it gets something wrong.

Outcome

A workable case for using AI grading responsibly, not a blanket yes or no.

Is human grading even the fair default?
  • Human grading gets treated as automatically fair, but tiredness, workload and unconscious bias all creep in.
  • AI grading is more consistent, but it comes with its own problems — it can be confidently wrong, hard to see inside, and can't actually be held responsible for a decision.
What's AI's actual role in grading?
  • The whole project is about this — a proposed setup where a human always has final say, but the system itself is constantly checked and open to challenge.
  • AI bias can actually be tested and audited in a way human bias can't — which reframes the real question from "which system makes fewer mistakes" to "whose mistakes are easier to catch and fix."
What would responsible AI grading require?
  • Put both systems side by side against the exact same standard — fairness, transparency, accountability, the ability to correct a mistake — instead of treating them as two separate debates.
  • Proposed four working principles: a human stays accountable, the system stays visible, it's monitored continuously, and students keep a genuine right to appeal.
What's this case built on?
Algorithmic-bias researchAI transparency & accountability researchUniversity grading-policy review
What are the working principles?
  • A workable case for using AI grading responsibly, not a blanket yes or no — a policy recommendation, not a deployed grading system.
Reflection
"The most interesting question about adopting AI is rarely about accuracy — it's about who's actually on the hook when it goes wrong."
Project Report

Read the complete report submitted for this project, including the research, analysis, strategy, and final recommendations.