(AI) Machine Scoring isn't the Problem: We are ALL Writing for Bots

I've been meaning to write about this for a while. But by now, many people have seen the scores and papers on STAAR exams. Very few essays have needed rescoring. Everything is going to be ok for both content scoring and mechanics. The scoring machine isn't the problem. Instructional relevance is. 

Yes - there are many things we still don't know about the language models used and how machine scoring actually works. That IS a problem but one of transparency from the agency, not for instructional implications.  

So let's think about how we can USE scoring realities to help make the process meaningful to those we serve. Here are five points that explain why we directly teach why machine scoring is relevant to daily life:  

1. All social media works with algorithms. You want to be seen and heard? You use terms and structures that the algorithm values. The scoring machine is looking for key terms from the text and prompt. 

2. Things like Netflix and Youtube give you content based on your behaviors. What you do changes what you are presented. You want to understand how to do the same for readers - even machines? Learn how to respond to reader needs with your text. 

3. Search engines work on what is promoted and searched and clicked frequently. Want a high score? Write what the rubric promotes and seeks. Want a high score? Use what's in the source text that fits the prompt. Scoring works like a giant Family Feud. Other students have already responded to this prompt with the source text and answer. The scoring machine "surveyed" a BUNCH of student essays and knows what most writers said and how the people scored their responses. Comprehend the text and the prompt. Use the text stuff in the written response, summarize how it fits into the text, and say in original thoughts about the connection to the ideas in the prompt. 

4. Search engines use key words to scan content and then matches those terms relevant to user's search terms. Want to be "seen" by the engine? Use terms that match what people want to see. For example, I've been writing for Sirius Education Solutions about thinking and reasoning in a project called Page to Pixel. There's a guy who reads over my stuff, restructures WHERE certain terms appear in the text, adds terms throughout, and chooses special search terms to embed in the code. I have no idea how the code works, but I can see that the key stuff goes in the intro, key connections to the topic are spread in each portion of the headings and body text, and the key stuff also goes into the conclusion. Sound familiar? We trace meaning and coherence through word choice and how we connect our topics through organizational structure. So, want a high score? Use key terms from the prompt and text in the intro and conclusion. Thread those ideas through each body paragraph with text that shares the same ideas/synonyms/textual context. The guy who edits my work for this puts these terms in bold so I can see how they thread through the text. I can see the connection and repetition of concepts through the entire text. 

5. Companies use machine scoring to preview resumes, cover letters, and communications. Priorities are given to text that includes what the company is looking for in the match of personnel to the job description and company goals. Understanding machine scoring is intensely personal and relevant to all those who are seeking employment. 

Machine scoring isn't new. Digital evaluation going on in the background of almost everything we do and see. And teaching without including explicit lessons on the relevance of these scoring machines isn't in the TEKS, but we'd be remiss if we didn't address the reality directly.