Reviewed against the exported Google Form (42 questions, 8 sections).
| Promised in the notice | Present in the form |
|---|---|
| “we ask you to create an anonymized ID so that we can link information from different survey points” | No such question exists |
| Right to withdraw consent and request deletion | Impossible without an ID; possible only if data are pseudonymous |
| Purpose: “teaching and university research” | Actual purpose is student council work |
| — | Google Forms = transfer to a third country, not mentioned |
Decide which model you want and make the notice match:
Send the final version to the LMU data protection office before fielding. Also confirm “Collect email addresses” is switched off in the form settings — the notice mentions the email display issue, but the setting is what actually governs it.
As exported, “I agree with the terms” carries the jump target (“Fahren Sie mit Frage 9 fort”) and “I do not agree” carries none. In Google Forms an option without an explicit target defaults to “continue to next section” — which would let people who declined proceed into the survey. Set “I do not agree” explicitly to Submit form (or a closing section that thanks them and ends), and re-test both paths on a phone before launch.
Study program × semester × gender × entrance qualification is identifying in small cohorts — M.Sc. Data Science (Elite program) especially. Commit in the intro to not reporting any subgroup below n = 5, and hold to it in the write-up.
Current form: open text only, stem asks about discrimination “because of your character/personality/person”, placed shortly after gender and origin, no support contacts.
Problems: free-text incident descriptions are de facto identifying in a small institute; “personality” is the wrong construct and will collect unusable noise; there is no closed item, so you cannot report a prevalence figure to the institute; and there is no signposting for anyone the question upsets.
Suggested replacement (three parts):
Have you experienced or witnessed discrimination or unfair disadvantage at the Institute of Statistics? - Yes, I have experienced it myself - Yes, I have witnessed it happen to others - Both - No - Prefer not to say
(filter: if experienced/witnessed/both) What was it related to? (multiple choice, optional) Gender · Origin, nationality or ethnicity · Language · Age · Disability or chronic illness · Sexual orientation or gender identity · Religion · Social or financial background · Something else · Prefer not to say
If you would like to tell us more, you can do so here (optional). Please don’t include names or details that could identify individuals, including yourself.
Support and reporting: [Ombudsstelle] · [Frauenbeauftragte] · [Psychosoziale Beratung des Studentenwerks] · [student council contact]
“No concerns at all” is a row inside a 5-point grid. Rated on the same scale as the other rows it is meaningless, and it will also be rated by people who did have concerns.
Fix: make it a filter first — “Did you have concerns before starting your studies here? Yes / No” — and show the grid only to those who say yes. That also shortens the form for a chunk of respondents and removes the 10×5 mobile grid problem in one move.
Same question: “Other” as a grid row with no accompanying text field produces uninterpretable data. Either drop it or move it to the open follow-up.
| Brief item | Status |
|---|---|
| How much students use AI | Missing — no frequency question anywhere |
| How students rate AI’s potential | Missing — no attitude items at all |
| Social media presence | Missing entirely |
| How to make StaDS@Work more attractive | Only as one line in a “value most” checkbox |
| Communication from the institute | Partly — two items, both double-barrelled (see 3.6) |
| Educational background (Bildungsstand) | Ambiguous — see 3.3 |
AI frequency — place before the tools question, and use it as a filter: > In the past four weeks, how often have you used AI tools for your studies? > (almost) daily · several times a week · about once a week · less often · not at all
AI attitudes (agreement matrix, 5 points, keep it to these six): - I check AI output before relying on it - AI helps me understand course material better - I know exactly what is and isn’t permitted in my modules - I would like clearer rules on AI use - My program prepares me adequately for working with AI - Assessment formats need to adapt to the existence of AI
Social media: > Which single channel would you most like to hear from us on? (single choice) > Instagram · WhatsApp channel · Discord · Email newsletter · LinkedIn · Website · Announcements in lectures · I’d rather not hear from you
Which content would be relevant to you? (max. 4) Deadlines and dates · Exam and study tips · Event announcements · Jobs and internships · What we’re achieving for you in committees · The people behind StaDS · Memes · Event recaps
StaDS@Work (shown to everyone, not just attendees): > Which formats would make StaDS@Work more appealing to you? (max. 3) > Short company talks · Hands-on workshops with real data · Speed-dating with recruiters · Alumni panel · CV check · Internship fair · Relaxed wrap-up with food and drinks · Hybrid or recorded
3.1 Semester (1, 2, 3, 4, 5, 6+) Fine for Bachelor, poor for Master — most Master students land in “5” or “6+” with no resolution. Consider 1 · 2–3 · 4–5 · 6–7 · 8+, and note that program (already asked) lets you interpret it.
3.2 Gender “Which gender do identify with?” — missing “you”. Options are fine.
3.3 “What is your highest degree of education?” If this is meant to capture social background (first-generation students), it doesn’t — a Bachelor student answering “Abitur” tells you nothing. If that’s the intent, add: Does at least one of your parents hold a university degree? Yes / No / Don’t know / Prefer not to say. If the intent is just prior qualification, the question is fine as is but note that “Ausbildung” and “Abitur” aren’t mutually exclusive — many people have both, and they’ll pick inconsistently. Consider making it multiple choice.
3.4 Current difficulties grid (4-point) No “does not apply” option. A native German speaker forced to answer the “Language barrier” row picks “No difficulties at all”, which then reads in your data as the language is fine rather than not applicable. Add a fifth column, or accept that the language and cultural rows are only interpretable within the non-German subgroup.
Also: the row list appears twice in the PDF export. That’s probably a print artifact of a wide grid, but open the live form and confirm it isn’t a genuine duplication.
3.5 Overall satisfaction Stem asks “how satisfied are you”, scale answers “Very happy → Very unhappy”. Align them — use very dissatisfied … very satisfied.
3.6 “Do you feel that you are missing out on information…? Do you ever feel like you don’t know where to find this information?” Two questions in one stem. Someone who feels well informed but can’t find things has no way to answer. Split into two agreement items: - Important information about teaching and events reaches me in time - When I need information, I know where to find it
3.7 “How do you get information about the institute?” The stem asks how (a channel question) but the scale answers how often (“Yes, all the time ← → No, not at all”). Reword the stem: “How often do you get information about the institute through each of these channels?” and use a proper labelled frequency scale rather than endpoint-only anchors.
Missing channels: WhatsApp/Discord groups among students, LSF/campus portal, noticeboards.
3.8 Attendance grid + “reasons not to attend” The reasons question is shown to everyone, including people who attend almost every lecture. Add a filter (show only if attendance < “often”), or add an explicit option “I attend nearly all lectures”.
Missing options: recordings, slides or notes are enough for me (probably the single most common real reason), room is too full / poor acoustics, long commute, care responsibilities, and an *Other: ___* field. “Illness/personal reasons” currently absorbs mental-health reasons, which you may want visible separately.
Typos: “Prefer studing on my own”, “Lecture is not helpfull”.
3.9 Old exams provided (0–100% slider) First-semester students can’t answer this meaningfully and the slider forces a value. Add “I haven’t studied long enough to say” or make it optional. Also: a slider produces false precision — nobody knows whether it was 40% or 55%. Bands (none · a few · about half · most · all) would be more honest and just as useful.
3.10 Exam preference (“What do you prefer?”) Binary with no midpoint and no “no preference” — you’ll force a choice from people who don’t have one. Make it a 5-point scale: clearly A · lean A · undecided · lean B · clearly B.
The option text also bundles three things (fewer exams + more ECTS + spanning multiple semesters). If a student likes fewer exams but not the multi-semester scope, the answer is unusable. Present the two models as a short framed block above the scale, and keep the wording symmetric.
3.11 Learning hours (two questions) Bottom category is “less than 25 hours” — a large share of respondents will sit in it, and you’ll learn nothing about the range that matters. Suggested bands: under 10 · 10–20 · 20–30 · 30–40 · 40–50 · over 50 · not sure.
Also, “including attending classes etc.” merges contact time with self-study, so you can’t tell a student who attends everything and does little else from one who attends nothing and studies alone — which is exactly the distinction your attendance section is trying to get at. Split into two rows (scheduled classes / independent study) if you can afford the space.
3.12 AI tools list No “None”, no “Other: ___“. Non-users have no valid answer and will either skip (leaving you unable to distinguish from missing data) or tick something they barely use. Missing tools worth adding: Microsoft Copilot, Perplexity, Mistral / Le Chat, DeepL or DeepL Write, NotebookLM, locally run open-source models. Typo:”Github” → “GitHub”.
3.13 AI access “Subscription provided by someone else (e.g. work)” merges employer, family and shared accounts. Split: free student plan from the provider · access provided by the university or institute · through my employer · shared account. The university-provided option matters most to you — it’s the one you can lobby for.
3.14 The two AI sliders Endpoint-only anchors (“much less ← → much more”, “beginner ← → expert”) give you a number nobody can interpret. Use labelled 5-point scales instead. Note also that the peer-comparison item measures self-perception and is subject to better-than-average bias — it’s interesting, but it is not a usage measure. That’s what the frequency question in section 2 is for.
3.15 “How would you rate the student council’s communication?” Three items only, all about events. Add: I know how to reach the student council and Communication is also available in English — the second one matters given you’re asking about German skills.
The row list also appears twice in the export; verify in the live form.
3.16 Overleaf question Buried near the end, and asks only about awareness. If the point is to justify what you spend money on, ask about the whole set in one grid: Overleaf Premium, past-exam archive, GOP-Tutorium, and whatever else you fund — awareness and use in one matrix (didn’t know it existed · know of it, don’t use it · use it occasionally · use it regularly). One question instead of several, and you get a ranking of what your budget actually buys.
3.17 “Which Student Council services and offers do you value most” Uncapped checkboxes — many people will tick everything and the ranking disappears. Add “select up to 3” to the stem. Google Forms can enforce this with response validation on checkbox questions; use it.
Unclosed parenthesis in “Additional study support (GOP-Tutorium, study material”. “Freshers’ / Freshmen” — pick one.
3.18 “What holds you back” / “What would convince you to participate” Both open text, both shown to everyone including people who answered “I’m already active”. Filter them, and convert the first to closed options with an open field after — these are the answers you’ll actually act on, and free text from 200 people is a lot of coding work for something a checkbox would capture: > Lack of time · I don’t know how to get started · I don’t know anyone there · I don’t know what the tasks would involve · The commitment seems too big · Language barriers · I feel like I wouldn’t fit in · Not interested
3.19 “Would you choose your current study program again?” Yes/No/Not sure is fine, but there’s no follow-up asking why not. The open question that follows is about program design, which isn’t the same thing. Add a filtered open field: “What would you choose instead, and why?”
Scale direction changes four times. Satisfaction runs positive-to-negative; difficulties run negative-to-positive; the communication grid runs disagree-to-agree; the information grid runs “yes, all the time” to “no, not at all”. Respondents build a mental model of your scales in the first few questions and then answer partly on autopilot — switching direction produces real measurement error. Pick one convention (conventionally negative on the left, positive on the right) and apply it everywhere.
Scale length changes too — 4-point, 5-point, sliders, and endpoint-only semantic differentials all appear. Standardise on 5-point labelled scales except where a slider genuinely fits.
No “can’t judge” option anywhere. People who have no basis for an answer will guess rather than leave it blank, and you can’t tell the two apart afterwards.
42 questions including several large grids is realistically 12–18 minutes, not the 10 people expect from a student survey. Expect meaningful dropout in the last two sections — which is where your student council questions live, i.e. the ones you most need answered.
Three options, in order of preference:
Add to the intro: estimated time, the n ≥ 5 promise, and a commitment to publish results by a specific date. The last one is the strongest single lever on response rate and makes your next survey easier to run.
Also consider a German version. You’re asking about German skills and offering the form only in English, which quietly filters your sample toward the confident — and the interface itself is already mixed (the branching label reads “Fahren Sie mit Frage 9 fort”).
| Current | Suggested |
|---|---|
| Which gender do identify with? | Which gender do you identify with? |
| Which semester of your current study program are you currently in (Fachsemester)? | Which semester are you currently in (Fachsemester)? |
| In general, how satisfied are you…? → “Very happy” | Align stem and scale (satisfied / dissatisfied) |
| What are reasons to not attend lectures | What are your reasons for not attending lectures? |
| Prefer studing on my own | Prefer studying on my own |
| Lecture is not helpfull | The lecture isn’t helpful |
| In how much of your courses… | In what share of your courses… |
| Github Copilot | GitHub Copilot |
| How do you have access to this tools? | How do you access these tools? |
| Have you been aware of the free Overleaf Premium Accounts? | Are you aware of the free Overleaf Premium accounts? |
| Additional study support (GOP-Tutorium, study material | …study material) |
| Which Student Council services and offers do you value most | …do you value most? (select up to 3) |
| Do you feel discriminated/disadvantaged because of your character/personality/person | See 1.4 |
Missing question marks on several items. A form with visible typos reads as less serious and measurably reduces both response rate and the care people put into open answers — worth one focused pass by someone who hasn’t been staring at it.