Taking part in the Greenroom study

Most community college faculty never get to practice a hard conversation with a student before having it for real. There is no coach in the room, no observer, and usually no budget. Greenroom is an attempt to give faculty a place to rehearse. Whether it actually helps is an open question, and this study is how I find out honestly instead of assuming.

I will publish the results so that other people who support faculty can use them, whatever they show.

Any instructional faculty member, full-time or part-time, at a college or university in the United States, who is 18 or older.

One exception: if you are in my reporting line, directly or through another manager, please do not enroll. That is a safeguard for you, not a judgment about you.

A few short questions before you join check this. If you do not qualify, nothing you answered is saved.

Answer five questions now about how certain you feel that you can handle specific kinds of conversations with students. About three minutes.

Use Greenroom however you want. There is no required number of rehearsals and no schedule, and nothing happens if you use it once and stop. The study is as interested in that as in anything else.

Answer the same five questions again twelve weeks after you join, or when the study period ends if that comes first, however much or little you have practiced. That time I also ask you to look back and rate how certain you felt when you joined, so I can compare two ways of measuring change. Then two optional open questions about what, if anything, you are doing differently, and what made Greenroom easier or harder to use. About five minutes.

I will email you when the follow-up opens and, if it is still unanswered, once more two weeks later. That is the most you will hear about it.

Optional: when the study period closes, if your follow-up was at least six weeks earlier, I will invite you to answer the five questions once more. About three minutes. Skipping it is fine.

You may skip any question.

Optional, and separate: at the end of any rehearsal you may share that one conversation. One or two readers, and for comparison an AI model, would read it for two things: whether the AI behaved as designed, and, in general terms, how faculty approach these conversations. You are never scored or reported individually. It is a separate decision every time, you can join and never share one, and I do not ask twice.

Optional: if what your dashboard shows does not match how the term actually felt, there is a box to say so. That mismatch is useful information.

The student you rehearse with, and the coach who writes your debrief, are built on large language models made by Anthropic, the same kind of technology behind chat assistants. There is no person on the other side, and no real student is involved.

When you rehearse, what you type is sent to Anthropic to generate the reply and comes back to you, with no name, email, or account detail attached. Anthropic does not use it to train its models and deletes it within 30 days, with two exceptions it publishes: if a conversation is flagged under its usage policy, Anthropic may keep it for up to two years (and its safety scores for up to seven), and it keeps what the law requires. Because some rehearsals involve students in distress, a few could be flagged. What is stored here is encrypted, as described below.

The models and the written instructions behind them are frozen for the length of the study, so that everyone is practicing with the same thing. If a fix is unavoidable, the change is recorded with the date and reported with the results.

These systems can be wrong, can miss the point, and can say something that lands badly. Whether they behave the way they are supposed to is one of the questions this study asks, which is why every session is checked automatically and why you may choose to share a conversation for a person to read.

Your answers to the questions above, and to any open boxes you choose to fill in.

The type of institution you teach at (community college, four-year, or other, never its name), how many rehearsals you had done before joining, and the background you gave during intake, if you gave it: employment status, discipline, and years teaching.

How you used Greenroom: rehearsals completed, which kinds of conversations you practiced, how long sessions lasted, whether you finished or stopped partway, which features you used, and simple counts drawn from the conversation itself, such as how many times you spoke and whether you opened with a question. These are numbers. No sentence you typed is stored in this record.

Whether the software behaved as designed in your sessions, such as whether the coaching cited its sources properly and whether a safety response was triggered. These are checks on the system, not on you.

Your rehearsal conversations, the coaching on them, and the term reflections Greenroom writes for you are encrypted on your device before they are saved. The key that unlocks them is held on the server that runs Greenroom so you can open your history on any device, which means this is not end-to-end encryption: someone with full administrative access to those systems could technically decrypt them. I do not read them, and there is no tool for browsing them. The only way I read a conversation is if you choose to share it, one conversation at a time.

Your name, email, and account details are never attached to study data. Study records carry a random code instead.

Nothing about your college, your department, or your employment goes anywhere. No college receives any information about you, and there is no report to any administrator and no institutional dashboard. That was never built.

No information about your actual students is wanted or kept. Please use roles rather than names when you practice; the practice partner steers you that way too. Scenarios you write have names and places removed automatically. A conversation you share has emails, phone numbers, ID numbers, and links removed automatically, and you confirm it contains no real names before it is shared.

Time: about ten minutes of questions in total, plus three if you choose the last optional set.

Discomfort: practicing hard conversations, and rating your own confidence, can bring up real situations that did not go well. You can stop a rehearsal at any point, skip any question, and leave the study entirely.

Re-identification, honestly stated: your study record carries a code rather than your name, and the list linking codes to accounts is held separately and restricted, so your data is coded rather than anonymous. Because the study is small, someone who already knew a lot about you might guess from a published combination of details. To limit that, I report only groups, never a group smaller than ten, and never describe an individual.

The usual risks of anything online. Data is encrypted in transit and at rest, but no system is perfect.

I do not anticipate any physical risk or any risk to your employment.

Practicing may help you. It may not. I honestly do not know yet, which is why the study exists, so please do not join expecting a guaranteed personal benefit.

The likelier benefit is to the faculty who come after you, and to the people who design professional development for them, who have very little evidence to work from about this kind of tool.

What you tell me also shapes the next version of Greenroom, which everyone gets, including people who never joined the study.

Once the study period has closed and you have answered everything it asks, your Growth page shows the averages of your own answers: when you joined, at the follow-up, and looking back. It is held back until then because seeing your first answers would sway your later ones. Nobody else sees them this way.

Greenroom is free whether or not you join, and it stays free for the length of this research. If that ever changed, you would be told, and nothing you had already contributed would be affected.

There is no payment for taking part. Participation is light and the tool is free, and paying people to answer questions about a free tool can put quiet pressure on those who could use the money. If you would rather be compensated for your time, the right answer is to not join, and to use Greenroom anyway.

You do not have to join, and you can use every part of Greenroom without joining, now and later.

Your decision has no effect on your employment, your evaluation, your contract or course assignments, your standing at any institution, or your relationship with anyone. No employer is told who joined, who declined, or who withdrew, and no supervisor is involved in any way.

If you are contingent or part-time faculty, that paragraph is written for you in particular. Nothing about this study can reach your job.

You can withdraw at any time, from this page, with no explanation needed. You tick a box to confirm, and press one button.

When you withdraw, I stop collecting study data immediately and delete what I already hold from you, except anything already included in a completed analysis or a published result. Once a finding is in a journal it cannot be pulled back out, and no consent form can honestly promise otherwise.

You are left out of every dataset created after that point.

Your Greenroom account and your own practice history are untouched. If you want your whole account deleted, you can do that at any time from your Account page.

Study records are stored under a random code. The list connecting codes to accounts is kept separately with restricted access.

Your rehearsal conversations, coaching, and term reflections stay encrypted and unread, as described above. What you type into the study's own boxes (the open questions and the note about your usage summary) is study data, stored under your code, with emails, phone numbers, ID numbers, and links removed.

When I look at study data, I see it through a view that shows the coded records only and cannot connect them back to a person. Those views are logged.

To be exact about it: I also run the system, so I have technical access to the database behind it, which is needed to keep Greenroom working and to carry out a withdrawal. That access could connect a code to an account. I do not use it to find out who is in the study.

I report only groups. Groups smaller than ten are left out rather than rounded off, and there is no public statistics page, now or in the future.

I am James Jarc, director of faculty development at a large community college in Cleveland. I was a dean and an associate dean before that, and before that a full-time faculty member and an overworked adjunct.

I built Greenroom myself, out of curiosity and a wish to keep learning, with the help of AI coding tools. I designed and developed it on my own time with my own resources, before bringing it into a research context. I want it to work as a resource for faculty growth. That is a real bias, and the study is designed to limit it: what counts as success is registered publicly in advance, the measurements are automatic rather than chosen after the fact, and a result showing no benefit will be published as a result showing no benefit.

I hold a leadership role at Cuyahoga Community College, which means I may hold institutional standing relative to some participants. That is why anyone in my reporting line is excluded, why my office does not promote the study or give credit for it, why no employer is ever told anything, and why participation is walled off from employment completely.

I do not make money from this. Greenroom is free and sells nothing. A paid version was explored earlier and never launched. This research is not expected to produce commercial profit. If that ever changed, participants would be told, and you would have no obligation of any kind.

Findings may be presented at conferences and submitted for publication in journals that serve faculty development.

After the study period closes, I will analyze the open-ended answers myself. Separately, as a comparison, an AI model from Anthropic will also analyze them, and will rate shared conversations, under the same terms described above and with no name, account detail, or study code attached. My own analysis is the study's result; the model's is only compared with it, and how that was done is reported.

Everything reported is grouped and de-identified. No participant is named, quoted identifiably, or described in a way that could single them out. A quotation would come only from text you chose to share, and only in a form that cannot be traced to you.

I may keep the de-identified dataset for later analysis, including future waves of this research. It will not be sold and will not be shared for marketing.

About the study or taking part, write to me at support@greenroompd.com.

About your rights as a research participant, or a concern you would rather not raise with me, contact Krystn Hood, Chair of the Cuyahoga Community College Institutional Review Board, at krystn.hood@tri-c.edu or 216-987-4777.

This study is about your teaching practice, not your wellbeing, and Greenroom is not a counseling service or a crisis resource. Nobody watches your conversations as they happen, so please do not use Greenroom to ask for help. If any of this brings up something heavier and you would like support, your institution's employee assistance program is a good first call, and in the United States the 988 Suicide and Crisis Lifeline is available by call or text at any hour. If you are worried about a student, your college's student support or care team is the right first call.