Month: January 2026

Precision in Assessment: Why Standardization Outperforms the Traditional “Curve”

In secondary and post-secondary education, teachers often face a “measurement gap.” This occurs when a highly rigorous assessment—such as a mock professional exam or a complex technical project—yields raw scores that accurately reflect performance benchmarks but fail to align with the broader institutional grading scale.

To bridge this gap, many educators rely on a “curve.” However, traditional curving often lacks statistical validity. Standardization, specifically through the use of Z-scores, offers a more mathematically sound and equitable alternative.

The Limitations of Common “Curves”

The term “curve” is frequently applied to two common but flawed methods:

  1. The Flat-Point Addition: Adding a set number of percentage points to every student. While “fair” in its uniformity, it does nothing to address the variance or “spread” of the scores.
  2. The Ceiling Curve: Adjusting the highest score to 100% and shifting others accordingly. This makes the entire class’s grades dependent on a single outlier, which can lead to volatile and inconsistent results.

These methods are essentially “band-aids” that fail to account for the relative performance of the cohort.

The Logic of Standardization (Z-Scores)

Standardization treats a set of scores as a distribution. By converting raw scores into Z-scores, we determine exactly how many standard deviations a student’s performance sits above or below the group mean.

The formula for calculating a Z-score is: z = (x – μ) / σ (Where x is the raw score, μ is the mean, and σ is the standard deviation.)

Once we have the Z-score, we can “re-map” it onto a target distribution (such as a school’s historical GPA mean). This ensures that a student who performs at the 90th percentile on a difficult assessment is rewarded with a grade that reflects that 90th-percentile standing in the gradebook.

Why Standardization is the Professional Choice

  • Maintains Rubric Integrity: Educators can grade with extreme rigor against high-level standards without fear of destroying a student’s GPA. The raw feedback remains honest, while the gradebook remains fair.
  • Corrects for Assessment Difficulty: Not every test is of equal difficulty. Standardization automatically adjusts for a test that was “too hard” or “too easy” by focusing on the student’s relative mastery within the cohort.
  • Statistical Defensibility: If a grade is challenged, the educator can point to a transparent, mathematical process based on the class distribution rather than an arbitrary “bump” in points.

By adopting standardization, we move away from “adjusting numbers” and toward “aligning distributions.” This practice respects the data produced by the assessment while ensuring that the final grade accurately reflects a student’s standing within the academic environment.

Innovation 2.0

The few who read this may have seen the post a while back called “Sunset“in which I reflected on the difficulties and, well, failures I suppose of trying to develop an LMS as a small business without a huge bankroll for a coding team and marketing. In 2007 when I started this and made some money from my inventions, the internet was very different.

So then AI came along. There is plenty of material for blog posts on how this transforms my teaching (I still teach remotely part-time). The big effort for me was trying to devise ways to prevent or at least make difficult the inappropriate use of AI by my students. Interestingly, I turned to AI to do this.

Like my colleagues who did not just surrender to AI student work submissions, I first worked on changing how I designed my assignments. That only goes so far.

Next I rolled up my sleeves and started tweaking my own code in this platform which I use for teaching remotely. Things like timers, detailed logging and response of student activity in a browser, hiding things until time has passed, and eventually on to getting an API key from OpenAI so that I could add a button that would analyze the logged data from student interactions on the platform and understand likelihood of inappropriate usage.

Once I started tweaking my old code, I noticed increasingly that the AI I was using to correct it, making enhancement above my coding ability, was itself increasingly having trouble with old-fashioned and out of date coding practices in the Perl language. I asked it about this. It explained that the code base I had (which is admittedly 20+ years old) was out of date such that it would not support a moderate customer base. The database itself, holding the work of myself and customers some going back twenty years, had obsolete features beyond the scope of this post to explain. The work to re-code and update this was enormous and overwhelming. That’s when the “Sunset” blog was written.

But then I had a cool idea for an application. I needed a way to let my AP French students practice and be evaluated asynchronously for conversation skills. I wanted to write this in a modern way using up-to-date code base. I used AI to write it. I was not as proficient in PHP as I was in Perl. I was tired of coding and wanted to focus on curriculum development.

The result was smashing! And from there I kept building… Three months later, I have nearly completed Innovation 2.0. Wow. I have moved from coding myself to directing the AI to to the detailed coding. I am now the creative director, no longer consulting programming language manuals or searching stackoverflow.

What’s especially exciting for me is that the new software works exactly as I wish it to. And it’s all in one place! That was why I started coding 30 years ago anyway! I like to build and invent.

So in January I will be using innovation 2.0 with my own students to refine and debug it and then move customers over in February and start offering this platform publicly. There are great new apps I can offer, a fully-integrated AI support system with guardrails and controls, effective live monitoring and more!