Every few weeks something changes here: new experiments, new settings, a better dashboard. This is where I write it down, newest first, for the people who run sessions rather than for developers.
Semester update
Two mails a year, nothing more
If you would rather not keep checking this page: before each semester starts, in February or March and in August or September, I send one short mail with everything that has been added since the last one. No marketing, no reminders, and a reply is enough to stop it.
The button opens your mail program with the subject filled in; you do not need to write anything, an empty mail to marian.sauter@uni-ulm.de is enough.
Your address ends up in my mailbox and nowhere else. Classperiment stores nothing, and it has nothing to do with your sessions or results.
What came when
·v0.21.1
A long day: feedback from the lecture halls, the platform round and this page
Two releases in one day. In the morning I built what colleagues had written to me after the November release, and four tasks learned something from it. In the evening came a round that brings no new experiment but makes the everyday work smoother. And then this page.
Line bisection: on a phone the thumb covered the line, so the first tap now places a marker you can move before you confirm. The old first-tap protocol stays available as a setting, for comparisons with earlier rounds.
The cancellation task has two new settings: a wide field, which asks phones for landscape, and a variant with closed circles as targets, after the Broken Hearts test.
Serial position curve: you can add a filler task, 30 seconds of counting backwards before recall, for everyone or for half of the class. With half the class, the dashboard lays both curves over each other and the recency effect disappears in front of everyone.
Sperling's partial report: after a colleague's question I measured the timing again, and it is right. New settings for the field duration, the answer time and the letters per row help a class that finds it too tight.
On the dashboard, "View as a participant" opens the running task in a new tab without counting as a participation. Handy on the lecture-hall PC.
The word lists of Categories and examples now follow the published norms.
Settings with choices are real drop-downs now instead of number fields where you had to know that 1 means on. The number is still what gets stored, so your lesson links keep working.
While you set up a session, a preview follows your settings, so you see beforehand what the class will see.
German dashboards finally write decimal commas and a proper minus sign. Mental rotation and the SNARC task draw a fitted line through the class means.
I ran all 88 tasks on very narrow phones. 23 of them reached a few pixels past the edge; now they fit exactly, and a phone narrower than 352 px gets a note before the start instead of a cut-off field.
Simulate class now mixes devices like a real lecture hall: about 70 % phones, 25 % laptops, 5 % tablets.
And this page. I wanted one place where you can read what has changed without digging through a changelog, plus the semester update by mail, see above.
·v0.20.8
Eight new tasks at once
The catalogue reaches 88 experiments. This batch is about perception and decision making: risky lotteries, the MouseLab room choice, the oddball task, inattentional blindness with the bouncing balls, prior entry, response priming, the SNARC effect and change blindness. All of them are marked Experimental, which means the paradigm is established but the classroom version has not been tried in a lecture yet. If you run one, I would love to hear how it went.
As with every new task, students see their own result only after you reveal the class result, so nobody spoils the effect for the neighbour.
Five new tasks, and the first changes that came straight from the lecture halls
Interference, retrieval-induced forgetting, the Wason selection task, the psychological refractory period and habituation join the catalogue, 80 experiments now. The more important part: colleagues wrote to me after their lectures about what had not worked in the room, and four tasks changed because of it.
Posner cueing and inhibition of return now show a triangle pointing up or down as the target, and you can set the interval between cue and target (and the cue's validity) yourself.
Anchoring lost its middle option "about the same"; now every comparison commits to a direction, as in the original.
Mental rotation shows the error rate per angle on the dashboard, not only the response times.
There is now a page that explains how you can support the project if you want to. Nothing is locked behind it, Classperiment stays free.
This was the big leap. In October I had asked what was missing in the lectures, and the answer was 51 experiments. Forty of them went live with this release, from memory and learning to judgement, evaluation and emotion, three of them whole new categories in the library. Alongside came things I had wanted for a long time: random assignment of students to versions of a task for between-subjects designs, results combined or by device, a roomier task display on laptops, and an Experimental label for everything that has not been tried in a lecture yet.
The task library has a category rail, a filter for core and experimental tasks and a preview of each task.
In every new task, students see their own result only after your reveal.
Colleagues had started sending me the paradigms they were missing, so the wish list now collects them by area, with their state: requested, planned, built. The first one built is the redundant target effect, which a colleague teaches in her research practical. And the texts of Posner cueing, inhibition of return and visual search now say more exactly what they measure, after another colleague's remarks.
Letter Search is a new task on location probability learning, and the first whose students see their own result only after your reveal. Card sorting was rebuilt closer to the classic sorting test, and the letter-match task along with it, after a colleague noticed that his class landed at chance. The header has a menu now, online samples can end with a completion link, and the tutorials page holds short videos for single steps.
·v0.12.2
A security review, task times and two data fixes
Before the code goes public I had it reviewed from the outside. The findings are fixed: class results now require the instructor key, publications contain class-level statistics only, and every submission is checked. More visible for you: every task shows how long it will take with the settings you chose. And two data fixes: the Flanker arrows no longer wrap on phones, and a condition without a single valid trial stays empty instead of reading 0 ms.
·v0.11.0
A page for instructors, with videos
If you want to show a colleague in five minutes what Classperiment does in a lecture: the new page for instructors has an overview video, the tutorial with chapters and transcript, and three cards on joining, the catalogue and the class result. The videos come from the project's own server, no third-party player, no tracking.
·v0.10.0
A feedback form on every page
Bug, idea or anything else: there is a small form in the footer, under the download row of the result page and in the dashboard's utilities. An email address is optional and only there so I can answer. The form never appears on a screen where a measurement is running.
·v0.9.1
New name: Classperiment
Cognimize Classroom became Classperiment, at classperiment.com. The old address keeps redirecting, so printed QR codes and the links in your slides still work. At the same time I started counting visits with a self-hosted analytics tool, without cookies, session codes or results, simply to know whether anyone is out there.
·v0.8.0
What the focus groups taught us
Students sat down with us and tried the tasks on their own phones. The result: every task now explains the touch controls they actually use, the practice round ends with a button instead of a timer, the result page explains its terms, and a class can compare itself with another class or with a benchmark. The catalogue can be tried without a session.