DreamApply’s human-centric view on AI use: highlighting the need for governed artificial intelligence in education

The integration of AI into higher education administration has moved past the question of "if" to "how." For admissions professionals, uncontrolled automation risks will become a multiplier of errors, not efficacy. DreamApply does not pretend to offer tools to replace admissions professionals, but instead focuses on a solid foundation for you to leverage industry leading AI models. This allows you to increase efficiency and exercise caution given AI’s nondeterministic nature.

Can AI deliver without context?

Unmanaged large language models (LLMs) are notorious for two critical failures in admissions: so-called hallucinations and unactionable output, especially when they lack context. To understand this, imagine asking for help reviewing your applications from a person who does not know anything about your specific admission process or the reasons why you ask for specific data. If you ask them to highlight items that require further consideration, they might spot outliers, trying hard to appear useful, but the review is likely to be unreliable for your purposes.

How do you ensure AI highlights what you want?

In DreamApply, users have access to the Application Highlighter tool. With it, you control how much data, which aspects of the application, and what highlights you want it to put out. To ensure it does what you want, begin by establishing a Contextual Layer or the context of each application field:

1. Why do we ask this question?

2. What are the applicant instructions and requirements?

3. What are the available choices?

4. Which choices are selected?

5. Which choices are not selected?

This allows the AI model to properly reason about the actual answer that the applicant has given, selected, or just as importantly, not selected. Context matters!

With this context, even the person we just imagined asking for help would do a much better job. Application Highlighter does the job faster, ensuring you use your time to review and make qualitative decisions.

Admissions professionals are meant to audit the highlights and make compassionate, qualitative decisions

More than 300 institutions in 40 countries have benefited from DreamApply in the last 15 years, and the company’s mission isn’t to trivialise admissions work by attempting to naively replace it with unbridled AI.

Through working with and for educational institutions, DreamApply software doesn’t strive for just speed, but accountability and actionability as well.

Institutions cannot delegate accountability to an algorithm. Our duty is to provide the guardrails that make human oversight possible, acknowledging the risks upfront which include:

• Data protection: Secure processing of essential data only.
• Ethical guardrails: The tool must not be used to encode or enforce discriminatory criteria, should strictly exercise evidence-based reasoning, avoid inferred, non-objective judgements. AI’s job is to process data, not assign value to it.
• Vigilance: Users must be educated that AI models are not deterministic or infallible and that human expertise is always required.

Admissions professionals will remain the drivers of this new technology. They will be empowered, have a reduced workload, and process much less data manually, but their insight, compassion, and strategic overview will ensure that the output is vetted based on institutional needs.

Experience, data, and structure of new AI technologies, including DreamApply’s Application Highlighter, indicates that the future of fair and efficient admissions lies in this sort of governed intelligence, not unbridled automation.

If you want to learn more about DreamApply’s features, request a personalised demo today.