Khanmigo, Mizou, and Packback each support one or two of the conditions well. However, none of the platforms supports them all. Most tellingly, metacognition is absent from all three: none asks students to examine their own reasoning, notice where it has changed, or explain why. What's missing isn't another feature. It's a single environment designed, from the ground up, around how critical thinking actually develops—one that connects these strengths into one intentional learning experience. 

What's missing is a single environment designed around how critical thinking actually develops, one that connects these strengths into one intentional learning experience. 

Above infographic created with Claude (Anthropic, 2024) based on prompts by the author and content from Abrami, 2015

It's tempting to treat critical thinking as a checklist of cognitive skills and to build an AI tool that simply drills those skills. According to philosophers Richard Paul and Linda Elder, to think critically is to hold oneself to intellectual standards while cultivating these intellectual traits:  intellectual humility, intellectual courage, intellectual empathy, intellectual autonomy, intellectual integrity, intellectual perseverance, confidence in reason, and fair-mindedness. (Paul & Elder as cited in Hughes, 2024)

This distinction has real consequences for how we design an AI tool. A system that only rewards correct answers can produce learners who are skilled but uncritical, good at manipulating information without the honesty to question their own assumptions or the empathy to take another's viewpoint seriously. Worse, an AI that hands over polished conclusions can quietly erode the very traits Paul and Elder prize, encouraging intellectual dependence rather than autonomy. 

Above infographic created with Claude (Anthropic, 2024) based on prompts by the author and content from Abrami, 2015, and Paul & Elder as cited in Hughes, 2024)

Designing to the expanded definition means building a tool that does more than assess reasoning. It models the intellectual standards in its own responses, prompts learners to examine their thinking rather than outsource it, and creates space for the traits—fairmindedness, intellectual courage, perseverance—to develop through practice. In an age where AI can generate an answer to almost anything, the more urgent educational goal is not faster answers but stronger thinkers. Paul and Elder's framework keeps that goal in view, ensuring the tool strengthens the habits of mind that make critical thinking a way of being rather than a procedure to be run.