Should Algorithms Decide? AI in Hiring, Admissions and Policing

Last spring my roommate Kai spent three weeks perfecting his résumé for a software internship at a tech company in Seattle. He tailored every bullet point, researched the company culture, even rewrote his cover letter five times. Two days after submitting his application, he received an automated rejection email. No interview, no phone screen, just a polite message thanking him for his interest. Later that month, a recruiter from the same company reached out on LinkedIn, impressed by his profile, completely unaware that their own system had already rejected him.

That moment stuck with me because it raised an uncomfortable question: if an algorithm can dismiss a qualified candidate before any human sees his application, what else are these systems deciding about our lives? I believe that while AI can assist in hiring, admissions, and policing decisions, it should never make final judgments without human oversight, because algorithms lack the context, empathy, and accountability that serious decisions about people's futures demand.

My first real encounter with algorithmic decision-making came during my own college application process two years ago. I applied to fifteen schools, and several used software to pre-screen applications based on test scores, GPA, and demographic data. At the time, I barely thought about it. The process felt efficient, almost neutral. But my high school guidance counselor, Ms. Rodriguez, mentioned during one of our meetings that these systems sometimes flagged students from certain zip codes differently, even when their qualifications were identical. She never said it was deliberate discrimination, just that the data these algorithms learned from reflected historical patterns of who had succeeded at particular institutions. That conversation planted a seed of doubt that has only grown as I have learned more about how AI operates in systems that shape opportunity.

The appeal of using algorithms in these high-stakes areas is obvious. Companies receive thousands of job applications for single positions, universities face record numbers of applicants every year, and police departments claim they need data-driven tools to allocate limited resources. AI promises efficiency, consistency, and the elimination of human bias. Proponents argue that algorithms do not get tired, do not play favorites, and can process information faster than any human team. In theory, a well-designed system could identify the best candidates for a job based purely on skills and experience, admit students based on genuine merit rather than personal connections, or predict where crimes are likely to occur without racial profiling. The technology itself is not inherently problematic. The issues arise in how it gets built, trained, and deployed, because every algorithm reflects the priorities and blind spots of the people who created it.

The problem I keep returning to is that these systems learn from historical data, and history is full of inequality. When an AI in hiring is trained on decades of employment records from a company that mostly hired men for engineering roles, it begins to associate maleness with success in those positions. Amazon discovered this exact issue when their recruiting algorithm started downgrading résumés that included the word "women's," as in "women's chess club captain." The system had not been programmed to discriminate, but it learned discrimination from the data it studied. In admissions, algorithms that weigh factors like neighborhood income or high school quality can reinforce existing advantages rather than level the playing field. In policing, predictive tools often direct officers toward neighborhoods that were already over-policed, creating a feedback loop where more surveillance leads to more arrests, which the algorithm interprets as confirmation that the area needs even more attention. My criminal justice professor, Dr. Patel, described this as "bias laundering," where subjective human prejudices get repackaged as objective technological outputs. The danger is that people trust numbers more than instincts, even when those numbers encode the same old problems.

I understand the counterargument that human decision-makers are also biased, perhaps even more so than algorithms. People have bad days, personal grudges, and unconscious preferences that affect their judgments. A hiring manager might reject a candidate because their name is hard to pronounce, or an admissions officer might favor applicants from their own alma mater. At least an algorithm applies the same criteria to everyone, and its logic can be audited and corrected in ways that private human thoughts cannot. Some of my classmates in my ethics seminar made this point last semester, and it is fair. We should not romanticize human judgment as if it is always wise and compassionate. But the solution is not to replace flawed human decisions with flawed algorithmic ones. The solution is to use AI as one input among many, a tool that can surface patterns or flag inconsistencies, while leaving final decisions to people who can ask questions the algorithm never considers: Why did this applicant take two years off between high school and college? What does this candidate's volunteer work reveal about their character? Is there context behind this arrest record that matters?

After watching Kai's experience and reflecting on my own path into college, I have become more cautious about where I let algorithms into my life. I still use AI tools for research and writing assistance, but I pay closer attention now to situations where software makes judgments about people rather than tasks. I check company career pages to see if they mention human review of applications, and I think twice before trusting risk scores or rankings that lack transparent methodology. What I cannot shake is the feeling that we are moving too fast, automating decisions before we have figured out how to build fairness into the systems themselves. Algorithms will keep spreading into hiring, admissions, policing, and beyond because they save time and money. But saving time is not the same as making better decisions. If we let efficiency become the only value that matters, we risk creating a world where people are reduced to data points, and second chances exist only for those lucky enough to fit the pattern an algorithm expects.

Did you like this example?

275

11