Posts

Quick note: I'm going to use AI synonymoulsy with large-language model applications through this whole post. They're not the same thing and artificial intelligence, as a field of study, has much more interesting applications that are generally ignored in the vernacular, particularly in education.

This morning, I attended some professional development that focused on artificial intelligence (ie, chatbots) in education. I ended up posting to Mastodon...a lot...

I've played both roles in this situation. I was an instructional technology coach for many years, so I understand the position the presenters and planners are in. I think, two years back in the classroom now, I have more complex thoughts on that time in my career that will show up in another post. This time, I was the classroom teacher who needed some presenters to tell me about the promise of AI.

The day was bookended with normal edtech hand waving, pretty much textbook - we started off with the "calculator moment" slide which reminded us all that calculators scared the hell out of teachers back in the day. That moved into a reminder that "AI is here to stay, so you're better off just using it" main talking point. It was tech-heavy, focusing on the tools and what they can do and the promise they hold to "change everything." The end of the day sounded like a chance to talk about more of the big picture implications, but those questions and points were either deflected ("We're not here to tell people what to do") or ignored as valid concerns for a teacher ("Yes, the environmental stuff is bad, but it helps with so many things.")

On the whole, sessions led by technology-first folks were all about efficiency, working more productively, and offloading the hard tasks like building assignments, valid assessments, and leaving thoughtful comments for students on report cards and IEPs.

Do not get me started on that last sentence.

All of the ideas shared were low impact and reduced the inflection points of human interaction we should have in our classrooms in the name of efficiency. If you can't talk to every student every day, have a chatbot tutor them til they get it right. Have students paste their story into the bot to get feedback instead of having students pair off and read each other's stories. At one point, when trying to demonstrate how easy it was to get started, a presenter said, "If you're good at talking with people, you'll be good talking with AI."

We go to school so we can work together. We build connections, we share ideas, we challenge...most of today glossed over what we lose as a result of diving in just so we're not "left behind."

All that said, my middle block was led by a split role collegaue. They teach and help with implementation. Their session began by acknowledging the difficult questions of ethical use and not just in the context of cheating. We talked about intellectual property theft, the damages caused to individuals and creatives, the damage to the environment...all of that before starting a discussion about what we've heard or seen with AI in the classroom.

Rather than assuming we need to be taught, they invided us to dialog before moving into a branching workshop where each of those concerns was addressed. There was not ulterior motive and they were prepared with complementary lines of inquiry. They planned it like a teacher and, while I don't agree 100% with their perspective, I felt invited to consider ideas.

I know I'm biased going in, but this was a real eye-opener for me since I've purposefully removed myself from edtech over the last couple of years. I'm discouraged by the overall tone of the day but I'm hoping that we can start to model what we want our students to do, especially when we're in group learning situations.

This year's stoichiometry test has come and gone with less than thrilling results. We're okay on unit conversions, but when we add that extra step of using the balanced reaction to describe ratios between substances, things fall apart.

This is a tough concept that I struggle to teach well. I don't know if more labs/hands on will help - I added one this year and I didn't see better results. Granted, we had several weather interruptions last week that preceeded the test, but even today, there are still large gaps in understanding that I need to do some digging into to figure out what is going on. My hunch tells me that is a "will" issue instead of "skill" issue for 90% of students.

There are many quotable sections in D'Arcy's retrospective on his career in education technology. A few in particular jumped out at me:

Context matters more in education than we initially assumed. A perfect learning object for one instructor might need significant adaptation for another. We learned that reusability often comes at the cost of effectiveness. And the term “Learning Object” became so genericized that it became meaningless.

The promise of having reusable content libraries, even between teachers at the same school, is compelling. But I'm often modifying or adjusting whatever I find, whether it's in the shared Google Drive folder with my colleagues or on the Internet. As they say, all politics are local, so is all teaching. I'm modifying - year over year - the materials I use with students. There is no one-size-fits-all collection of stuff to use with teaching.

But the real story was more complex. [Learning Management Systems] tend to shape pedagogy in subtle ways. Their design implies a certain model of teaching and learning - one based largely on content delivery and assessment.

I had the great fortune of learning to teach from a wonderful mentor who took advantage of early digital learning tools to make it more accessible for students. My first position also encouraged us to develop strategies for using wikis and blogs. I felt frustrated by the push to move into an LMS for coursework because I already had systems and strategies in place for those kinds of interactions.

In the next paragraph, D'Arcy concedes, "That said, LMSs solved real problems. They made online teaching accessible to many more faculty." This is on point because the burden of online teaching tools, at the time, was heavily skewed toward the non-technical teacher. I was comfortable trying new things and figuring systems out. Others either aren't comfortable or don't have the mental energy to do that prep work, which is also fine. Years ago, I felt like the LMS could be used as a stepping stone into other tools that do some things better. And that's a good thing.

If you're in education, go take time to read this.

Where Part 1 ended the story of Paul, Duke of House Atreides, and began the story of Paul, "Muad'Dib," Part 2 ends the story of the Lisan al Gaib and begins the story of a would-be God Emperor’s holy war that he’s not exactly psyched for because Chani’s right to call him out on his bullshit.

This is great take on why Dune Part Two is so powerful. It's not "true to source," but it captures the tragedy playing out through Paul. I think the film adaptations cut through all the extra stuff in the book that couldn't have made it in anyways.

In my mind, the Villeneuve version of Dune is right up there with Peter Jackson's Lord of the Rings adaptations.

Something came over me this weekend and I decided to watch a bunch of movies. We had a long weekend for the Dr. Martin Luther King holiday, so Friday night, I decided to do a rewatch of Dune, Part One. I missed the theater run and watched it on one of the Big Streamers at some point. The rewatch was just as good and, after having finished the book again in 2024, I think the movie holds up really well.

Saturday night, I decided to watch Godzilla Minus One. I remember getting the original Godzilla movies growing up from the public library, so there was a bunch of nostalgia in this for me. I watched it in Japanese with subtitles turned on and it was definitely the way to go. The acting was amazing and I was fully bought into the story.

Sunday night, I re-watched Dune, Part Two. I saw this one in theaters back in April (that was an experience) and I wanted to watch it in closer proximity to finishing Part One. I had forgotten about how well the second part shows Paul's descent and I think the conflict plays out exceptionally well through Chani's story. The ending left me feeling really sad because there is no real path to redemption. I also read Dune Messiah last summer and I'm curious how that particular story will make its way into the (rumored?) third movie.

The artic air settled over us on Monday night and school was cancelled on Tuesday. Feeling melancholy over the other events of January 20th, I decided to watch Oblivion because I had a vague memory of seeing it, but not remembering any of the story. It was a good escape and, pleasently, I did not remember the main story twist, so it was fun.

This is the most I've watched in such a short period in quite a long time, but it was fun to just watch the stories play out. On the other hand, they started to put me in a mood and I've had a hard time focusing on other priorities, so I think that will be it for this time around. Maybe next year, I'll do it all again.

We're neck deep in unit conversions in chemistry and we've come to that bridge, yet again, where we have to wrestle with stoichiometry. Converting between units has gone better this year because I introduced with some models using common units like dozens, grosses, scores, and others. They counted beans, did their multiples and partials, and were able to get to the mol (I'll always leave the "e" off) without too much of a problem.

But then, we got to chemical compound conversions using balanced equations. Every year, this is where we come to an impasse for some students.

In the past, I've used the "mol island" metaphor and associated diagram to help make sense of the conversions that are happening to get from A to B:

Each year, moving from mols of A to mols of B is the hardest part, by far. The idea that you can convert between substances is hard to capture and, really, I should probably develop a new lab to help with the concept, but this year, I don't have time to at the moment.

This year, they're also struggling with procedure recognition - understanding where they are in a given process compared to where they need to go. I started tinkering with the idea of a flowchart, similar to how the old Holt Mini-guide to Problem Solving book models it:

This is nearly the same as my version, just less circley. I needed to do something different, and I ended up with this version:

I wasn't really planning on using it with students this year, but on a whim, I threw it on the board today and most of the class immediately asked for one to put in their notebooks.

I'm not sure what it is, but the yes/no questions along the way help them with the step-by-step processes needed to make it through a conversion, more than the island chart did. I think the connection between the questions and following the path to finish is providing an almost tactile interaction that is helping link ideas.

It's definitely a crutch, but I've always been more interested in teaching processes over fact retention. I think this is a better tool to help them approach novel situations and, with repeated use, they'll start to rely on the diagram less and less.

Frankly, this was born out of frustration and is more or less a happy accident. I'm hoping results actually pan out when I start to assess the skill formally next week.

I've struggled to efficiently grade web dev projects this year. Github Classroom is kind of a pain becuase I need to go to each repo individually and students are struggling with the git workflow. I decided to take that out for now. The next best method is to have them zip project files and then submit the zipped archive that I can extract. It was still a lot of clicking between folders.

Then, I realized (finally) that I can just extract the directories into one parent folder and run Python's built-in HTTP server. It makes clicking around the projects much easier and I can pop open the source tab right in the browser. No more opening and closing files between apps.

I wish I'd thought of this back in September.

I don't normally do blog challenges, but I got pinged by ~hyde with the call, so here goes:

Why did you start blogging in the first place?

I don't remember a specific thing that got me started. My first post talked about students and learning and technology, so it was probably something I heard during a professional development at school. This was my first year teaching and the school was pretty technology-forward, so I was doing a lot of exploring. That early writing is pretty cringy as I read it 15 years later...

My writing really beagin in earnest in 2011 after attending an education technology conference and getting connected to other teachers through social media and blogs.

What platform are you using to manage your blog and why do you use it?

The current version of this site is built with Pelican, a Python static site generator. I've been using this for a little over a year now becuase I moved from a shared hosting provider to a Linode VPS and I needed to cut down on resources. I tried WordPress at first, which is what I had been using for many years, but this new host only has 1GB of RAM and a static site made more sense. I also don't do enough shared writing or management to really necessitate a full CMS.

Have you blogged on other platforms before?

I had a brief stint of running my site with Jekyll in 2018. I also had it running on Anchor CMS a few years before that. I still miss Anchor.

How do you write your posts?

I have a private git repo that I use to sync posts between machines. Pelican turns Markdown files into HTML when the site is generated. I use Helix to write and git to push to the repo. A pre-receive hook on the server extracts the new files and rebuilds the site on each push.

When do you feel most inspired to write?

It depends a little on the week. I've done some work to make short posts easier to publish and I'm trying to make that more the norm. I tend to think about teaching the most because it's what I spend most of my day doing and I'll often chew on those posts for a couple days before actually starting to type. I don't know that I ever have a "fully finished" idea before publishing because this site works best when I use to to process my thinking rather than "publishing" my thinking.

Do you normally publish immediately after writing, or do you let it simmer a bit?

I tend to publish right away. If I have a change of heart, do more thinking, or expand my ideas, I'll either edit the post and note the update or I'll just write a followup post. I don't keep drafts around because if it's not enough of an idea to write it out, I don't need to worry about creating a backlog for myself.

From time to time, I will mark a post as "scheduled" to publish at a later date, but that's pretty rare. More of a nice to be able to do than something I reach for.

What's your favorite post on your blog?

Alan Levine asked a very similar question on his blog, asking which post was a contender for best post. I read it in November when he published it and the tab is still open on my phone becuase I will, someday, go back and comment.

I don't really know. Maybe this one?

I think some of my best writing is the life stuff. Not so much the thoughts on schools or tech, but the post about the time I had to fix a bad car problem or the other time I had to fix a car problem.

I don't think I can pick a single best post, but I think I can say, 15 years, in, that my Life category is my favorite collection now.

Any future plans for the blog?

I've got a couple things I'd like to do, like auto posting new articles to other places and making this the first place of sharing rather than jumping to social media. I want to write more about life stuff and tell more stories of living simply.

Who will participate next?

My fedi circle is pretty small still, but I'd like to see a post from Chris, Alan, and Tom mainly becuase I always make sure to read what they write when it pops up in my reader.

My students were incredible with the iodine clock lab. Given the complexity of the procedure and the accuracy with which they needed to measure, they rose to the occasion.

Everyone split into their groups and immediately read the materials lists and read the procedure before even getting chemicals. They measured stock, asked good questions ("Is tap water okay?") and delegated tasks.

The soft click of glassware and hush of partners cooperating was musical.

I wrestled hard with an iodine clock reaction this week. I haven't taught kinetics in many years, so I was re-learning a lot as I went. I'm still confused a little by the data I got vs what I expected, but I think it's a case of the real world being much messier than practice problems.

D'Arcy manages to say in two sentences the conflict I feel over using an LLM:

If the resource and environmental impact problems were solved, would it suddenly become ethical to use LLMs? If the ethical issues around intellectual property and training of LLMs were solved, would I still be this conflicted? I desperately want a simple answer to all of this, and know that it’s just not possible.

These are my main two holdups, but even if they weren't problems, would I feel the same? I'm not sure.

Simon Willison has a detailed overview of major changes in large-language models from 2024 that I took time to read today. I'm a skeptic, especially because of the copyright and environmental issues that come with creating and running these services at scale. Some items that jumped out:

The really impressive thing about DeepSeek v3 is the training cost. The model was trained on 2,788,000 H800 GPU hours at an estimated cost of $5,576,000. Llama 3.1 405B trained 30,840,000 GPU hours—11x that used by DeepSeek v3, for a model that benchmarks slightly worse.

I don't pretend to understand the complexities of the models and the relationships they're trained to form, but the fact that powerful models can be trained for a reasonable amount (compared to OpenAI raising 6.6 billion dollars to do some of the same work) is interesting. Costs are down, which means that electric use is also going down, which is good. Simon makes the same observation in his section on environmental impacts:

A welcome result of the increased efficiency of the models—both the hosted ones and the ones I can run locally—is that the energy usage and environmental impact of running a prompt has dropped enormously over the past couple of years.

An interesting point of comparison here could be the way railways rolled out around the world in the 1800s. Constructing these required enormous investments and had a massive environmental impact, and many of the lines that were built turned out to be unnecessary—sometimes multiple lines from different companies serving the exact same routes!

The resulting bubbles contributed to several financial crashes, see Wikipedia for Panic of 1873, Panic of 1893, Panic of 1901 and the UK’s Railway Mania. They left us with a lot of useful infrastructure and a great deal of bankruptcies and environmental damage.

I think the last paragraph is where I'm still sticking. Companies are plowing ahead to give us AI in things we don't want it in at the cost of A) an economic bubble and B) massive environmental damage. I'm seeing economic impacts near home with datacenters being built at massive tax discounts which benefits the corporations at the expense of residents. There will be bills to pay and right now it doesn't look like it'll be corporations.

I'm not going to start using an LLM daily, but reading Simon over the last year is helping me think critically. I dabbled with self-hosted models, which was interesting but ultimately not really worth the effort on my lower-end machine. Maybe that will change as systems become more and more optimized for more general use.

If you have time, I would encourage you to go read the entire post.

One of my kids said tonight that they want to have 84 children. Another one of the kids exclaimed, "It'll take, like, five million dollars to feed them." That got me wondering if my wife and I have spent one million dollars in our time together.

Our rough estimate using our general food budget, mortgage payment, and house expenses (electricity, gas, and taxes) would take us 40 years to spend one million dollars. We live modestly, but that number was higher than I expected, honestly.

It reminded me of the "one billion dollars" on MS notepad video (lots of NSFW language, so be careful). The fact that there are people with multiple hundreds of this kind of money is nauseating.

I've made some small tweaks to the blog to kick off the new year. For starters, I made the layout slightly narrower because the original theme took up a ton of horizontal space. I don't do any full-bleed layout elements because of the two-column layout, so this will help keep things nice a neat.

The biggest change is how I'm laying out posts on the main index. Instead of just titles, I'm now showing small excerpts. Jinja templating allows me to pull the text to a certain point and then add the "Read more" link when needed. I also did a little bit of preliminary work for posts that I want to throw in that don't necessarily fit the "traditional" blog post format with a title.

Sidebar: it is wild to me that a blog post has a "traditional" format now.

Lastly, in an attempt to have a more semantic layout, I've added a proper footer element into the article element for the post data. I've included a permalink, a server-rendered word count, and then the category as a link. This will let me remove a little bit of Javascript from the post templates I was using to calculate the word count.

We'll see if these changes stick.

Jack Baty had a blurb on his blog this week about filtering server logs to use with GoAccess. I have logs running on my various sites but I don't go through the logs too often, mainly because I don't know enough about tooling to make them more readable. I had tried AWStats a year or two ago but ended up dropping it because it made me wonder why people didn't come to my site. It was also a pain to set up.

It's 2025 now and I decided to throw stats back in, but not as a live service. GoAccess was also much easier to get up and running, so thanks Jack for linking to it in your post.

The TUI was nice and I could pop open a log file to look at the overall stats on one of my various domains. I decided to clear out old gzipped log files and rework my logging strategy a little bit. Nginx will rotate the log daily and put each subdomain into it's own directory that I can analyze whenever I want to. I'm keeping those log files for 30 days at a time because I don't get that much traffic and I don't think my disk space will be filled up.

I followed a wonderfully detailed tutorial by Arnaud Rebillout to have a static HTML file generated. It isn't hosted anywhere and I can rsync it down to my machine to look over every now and then. This method has a small persistent database file which prevents duplicate entries, which is nice.

Will I keep it around? Maybe. At least I don't have to do anything manually to generate the report. It'll be something to pop open when curiosity bites rather than a chore to keep up with.

2024 is nearly finished and one of my last things to do is think over my goals for this year. One year ago, I wrote down some things I was hoping to do. I checked in on those in July and now it's time to look everything over.

• Read 40 books - I didn't meet this one, but I read 29. Having some abritrary number of books to read helped me continue to read regularly, so I'll do that again. I don't know if I'll take the achievable approach or set some ridiculous number just to have a number to shoot for.
• Add blog comments - Still available at the end of every post.
• Cut phone use - This was probably my worst of all. I didn't really keep track of stats like I did with some of these others, and that's probably why it was harder. Sometimes, I just don't have the mental capacity to read or do much other than scroll. It's not a great habit and I need to find another outlet for those fatigue-induced scroll sessions. I'll probably keep this as a goal for next year, but I need some clear cut indicators I can work on individually.
• Ride my bike 150 miles - I crushed this goal. Turns out, I really like riding my bike. I logged nearly triple my goal - 430 miles on a blue 1986 Fuji Allegro - including a wonderful 15th anniversary ride through Chicago and a couple rides to Lake Michigan. We bought an indoor spin bike to keep our cycling muscles exercised during the winter and I'm looking forward to extending my cycling next year. My rides are logged in Strava if you'd like to connect there.
• Get two new courses approved - I am deep in the midst of teaching both Advanced Chemistry and Computer Science (web dev) as new courses this year along with my normal chemistry courses. Three preps is a ton of work - I'm really glad I'm teaching these new classes, but I will be cutting it back to one of those electives each year rather than both.

Every break, we decide to tackle some home improvement projects. This week, we decided to finish putting drywall in the basement. We've been half done for about a year, so this past fall, I finished the framing so we could wrap this project up.

As I was working, I put in new electrical outlets using a coil I had left hanging from the last drywall phase. I wired everything up and then went ahead and tested the connection.

No power.

sweaty face emoji

That's right, I'd not hooked up that wire so that I didn't electrocute myself. Past me was safety conscious.

So, I opened up the first outlet in the run and hooked up the loose wire. Happy to have remembered my foresight, I turned everything back on only to find...no power.

two sweaty face emojis

In all this, the south wall had no outlets, which was strange. I wouldn't have left an entire wall without power, so we went hunting. The electrical boxes behind the drywall would've left a slight bulge. I grabbed my level and, sure enough, there was a subtle bulge in the wall. But, how high was the box?

In such cases, I scratch my head and look at the floor.

Past me was even smarter than I gave him credit for. Knowing, for some reason, that I couldn't cut out the outlet boxes I had put in the wall that I would need to find them, I had marked the floor with the stud and the measurement to the center of the box. I punched and cut with my drywall saw to find a box with wires ready to go.

I spent 20 minutes getting those boxes wired up and, hey presto, the entire basement has power. No cutting or removing drywall hung last year because past Brian was a smart dude.

This phase went so smoothly that my father in law offered to come up and help us get the first coat of plaster on the seams.

Breaks aren't always this productive, but because of some good thinking done by yours truly in the past (that I honestly don't remember doing, so don't give me too much credit), we were able to knock out a relatively big job in just a few hours of work.

Well, it's the twelfth month of the year, which means this is my twelfth reading log blog post. I did not hit 40 books this year, but I did manage to finish 29, so nearly 75% completion. I didn't really keep track last year, so I don't know how much more I read, but I did read much, much more than 2023.

In December, I read:

Revelation Space - Alastair Reynolds

It took me two tries to read this one. I will not be continuing in the series. The first third of the book felt confusing, with different characters in different times dropped right into an entire universe that doesn't get a whole lot of explanation later in the book. Humanity has spread to new stars and has split into different factions with their own norms and histories. There are also interactions with alien species, but again, lacking any connection to the plot unfolding.

Several of the characters are there without much reason and several don't reveal any motivations for their actions. The second third was better than the first, but the ending felt flat. The conflict plays out and some questions are answered, but I'm not convinced the book needed over 500 pages to end where it did.

Accidental Astronomy - Chris Linott

This is brand new and happened to be on the "New Nonfiction" shelf at the library last time I went. I can't really focus on nonfiction during the school year so I try to cram some learning in during breaks.

A wonderfully written, easy read about some of the weird discoveries in astronomy made completely by accident. Linott's writing style is casual with enough technical detail balanced out with clear narrative and witty footnotes on almost every page. I enjoyed this book immensely.

As we roll into January, I will start my count again. Here's to the 2025 reading log!

All of my reading is on LibraryThing if you want to take a look at everything I read in one place.

My hobby project is a self-hosted grading/feedback platform for students to track progress on learning standards. I call it OutcomeMaps, mainly because standardmaps.com was already taken, and it is probably the second most-used site on my machine each day.

Everything we do in chemistry is linked to a speciifc learning outcome - usually a skill based on theory (describe the structure of the atom) or on computation (calculate the mass of product expected from these inputs). Every unit has two or three defined, each unit building on the previous, and students are graded on their proficiency on these standards. The problem is that I didn't have a great way to track progress on their learning.

So, I set out to build a little frontend for a database. It's simple in the sense that students are essentially read-only users and they can see a catalog of all feedback they've received through the year and a dashboard of their progress on any given standard. As the teacher, I can create/edit/score anything, but the main interaction layer is feedback on the skills.

The key is that it helps me make decisions and helps students see their opportunities for improvement. I don't think there's one way to quantify "proficient" but this system has helped me collect data which can then be used for plan for instruction and address emerging needs.

One thing I changed this year is to give binary scores to students. I started thinking about this switch back in March, mainly becuase my grading scheme at the time had areas of uncertainty in the scale that I wanted to eliminate. On paper, this turned into a 0-1-2 scale: no submission, improving, demonstrated. Students have been trained to glance at the number but, more importantly, pay attention to the written feedback. Based loosely on the single-point rubric, it does a better job of getting students to think about their thinking and not just about task completion.

On the web, I turn that into a "check or x" - the UI doesn't actually show their score, just a checkmark or an "x" to note proficiency on that thing. The feedback is displayed right alongside each assignment's result so they can identify their own strengths and areas of improvements for each standard. The goal is to reach a check by the time the unit test rolls around.

This year, I built out an admin panel with some simple charts that will help me - at a glance - look for patterns in each course.

At this point, I was calculating proficiency on the fly with each request because it was just one roster at a time (~25 students per class). This chart needed to generate three objets with data for each standard, so I was looping over each student multiple times. The chart took several seconds to load, which was less than ideal.

I did some work on the database to create a new association table that would record when a student was proficient - generally after taking a test - and I could store that record permanently. I also wrote in a method to override proficiency manually - the use case being that a student might come after school and do some one-on-one work that wasn't on a specific assignment. I can store that proficiency with a click instead of creating a new assignment for that one thing.

This let me query all of those records at once and cut down the load time by orders of magnitude. It also allowed me to set up per-standard comparison charts in the admin panel:

Now, as we're in the unit, I can look at the progress for each class on one screen. This used to be manual - I would look class by class and think over next steps. Now, I can reflect more on what I do differently from period to period and how that translates into success on the skills.

This year pushed me in terms of my coding ability. I've never been comfortable with database joins, but I'm using them all over the place now. I also have a better handle on how to set up models to get the data I want. I'm also coming up with ideas about how to help shepherd students through their own reflections on learning. One goal for this next year is to introduce tagging on standards ("theory" vs "computation", for example) which can help target even more interventions and study strategies for students.

Beyond that, I'm more strategic about my CSS and template structures. I've done a lot of maintenance work to clean up templates and I'm getting ready to tackle a full CSS cleanup (it's a hot steamy mess right now). It's by far my most long-lived project in terms of raw code written to do a thing (as opposed to raw words here) and I'm really proud of how far it has come from my really bad first couple of versions.

Year 15 in teaching, another post on grading. I'm still using standards based grading and I'm still thinking (overthinking?) how to represent student progress on each standard.

I've gone through several scoring style changes. Most recently, I've switched from my old four-point scale to a two-point representation, mostly based on an old post from Kelly O'Shea on her SBG system. Students are given a 0-1-2 score on any skill we work on in class. A 2 means they've demonstrated the skill successfully. A 1 means they've got some work to do and a 0 is reserved for "No evidence." It's simplified my scoring and I think students have a better overall understanding of what "showing proficiency" means in practice.

My deeper struggle is how to represent the final result to students. Each of the feedbacks they get through a unit points them to in-progress fixes. In other words, small errors in their processes get caught and they're able to fix those things on the next attempt. They can see the linear growth from 1's on assignments to 2's. But what about at the end of the unit? How is all of this wrapped up?

Like Kelly, I've decided that tests are the "hard limit" check. If it's not perfect on the test, then it does not show proficiency. All semester, I've been putting that standard score is more or less based on the test attempt. So, even if they've been proficient up until that point, mistakes on a test will cause me to hold off putting that last binary yes/no on the skill into the book as completed.

Kelly's rationale is that the test should be their best true demonstration of understanding, and I agree with that mindset. It's a controlled environment in which they're asked to do the thing correctly. This requirement also helps reinforce the fact that copying and/or otherewise doing unauthentic work leading up to tests is a waste of time for the student becuase the still have to be able to perform the skill at the end of the day.

To help collect evidence over time, I've got a little web app students can use to see their progress on standards. All feedback goes on their papers but it also goes into this app so they can watch their progress throughout the unit. A student may be showing proficiency on a skill throughout the practice work only to be hit with a "not proficient" based on that test. I'm trying to reconcile that story in my mind so that it makes the most sense for them.

The message I need students to really internalize is that learning is a process. Tests are a way for me, the teacher resposible for fostering and reporting their growth in the content, to check that growth reliably and with validity. Reporting their standard proficiency based on the test is the best true check. Their growth up until that point is their indicator for what to fix leading into that final check in the normal flow of the class.