Equity Lens

Working Group on Equal Opportunities of the German Physical Society

Equitable decision support for selection committees

Start or join a session

The chair opens a session and sends one private link to each committee member. Members open their link on their own device.

What kind of selection is this?*

Sets which criteria are collected and which checks run.

How many members will vote?*

Results stay sealed until this many ballots are in.

Do not offer this session for research or AI training

Nothing is ever sent automatically. At the end of a session you may choose to contribute an anonymous summary — coarse buckets only, no names, no institutions, no link back to this session — to AKC's open dataset. Ticking this box removes that option entirely, so it cannot be pressed by accident.

Send one link to each member

Each link works once. Hand one seat to each person — by private message, or by letting them scan its code. Anyone holding a link can cast that seat's ballot, so show one code at a time rather than projecting them all.

Checks in use

Selection type: —

Always on

Gender balance, institutional concentration, career interruptions and impact beyond citations. These run for every selection type and have no switch.

Optional for this selection type

Anything switched off here is recorded alongside the session, so the open dataset shows which checks were active rather than quietly omitting them.

Candidates

Candidate info

Name*

Career stage*

Institution*

Gender*

Used only to compute the gender balance flag.

Academic impact

h-index*

h-index = h papers with at least h citations each. Scopus, Web of Science and Google Scholar give different values for the same person, so use one source for every candidate in the pool. Full Scopus profiles need an institutional subscription; without one you will see a limited preview.

Year of first publication*

Prior invitations (last 3 years)*

m-quotient = —

Key publications (max 3, one per line)

Impact beyond citations

Patent / technology transfer Policy citation Major media coverage Major award / prize

Career stage considerations

Career interruption of 6 months or more Non-traditional career path

For example parental leave, caregiving, health conditions, military or civil service. Only the fact of an interruption is recorded, never the reason — which is why this tool holds no health data about anyone.

Bias check

This tool does not rank candidates or recommend an outcome. It makes patterns visible so the committee can apply its own judgement.

Open the vote

Publishing encrypts the candidate list in this browser and sends it to members' devices. The server stores only ciphertext and cannot read it.

Double-blind: hide names and gender from members

Candidates appear as "Candidate 1", "Candidate 2" in a shuffled order, and key publications are withheld since a title identifies the author. Institution, career stage and metrics stay visible. You still see everything, so this reduces bias among the other members, not your own. Decide before publishing.

Cast your ballot

Your ballot is counted but not recorded against your name. No one, including the chair, can see how you voted.

Ballot box

Checking…

Contribute to the open dataset

AKC is building an open record of how selection committees actually decide, so committees can see where the field is going and where patterns repeat. Contributing is entirely optional and nothing happens unless you press the button.

The record below contains no names, no institutions, no exact counts and no dates — only coarse buckets. It carries no link back to this session or to any ballot. It is anonymous, which is why it can be published openly and why no consent from candidates is required. Read it before you send it.

What kind of selection was this?*

Field*

Exactly what would be sent

Choose a selection type and field to see the record.

What this tool stores, and where

Stays on the chair's device

Candidate names, institutions, career stages, gender, metrics and key publications are typed into the chair's browser and held there. Before the list is sent to committee members it is encrypted in that browser with AES-256-GCM. The key travels to members inside their seat link, in the part after the #, which browsers never transmit to a server. The server therefore holds ciphertext it has no key for.

Held on the server during a session

The encrypted candidate list, unreadable without the key.
A one-way hash of each seat token, and whether that seat has voted.
A count per candidate identifier. The identifiers are random strings, not names.

These sit in two separate tables with no column joining them, which is why no one — including the chair, and including whoever operates the server — can determine how any member voted. That is a property of the schema, not a promise.

Where it physically sits

On Cloudflare Workers, in a storage object pinned to the European Union, so processing and storage stay in the region. The object's random identifier is logged outside that region for billing and diagnostics; it contains no personal data.

How long

Until the chair presses "Delete all session data", which erases everything for that session immediately. Automatic deletion once voting ends is not yet implemented and is the next change planned. The chair's own working copy lives in the browser tab and disappears when that tab is closed.

What leaves, and only if asked

Nothing is transmitted to AKC's open dataset unless a chair explicitly chooses to contribute, and the exact record is shown before it is sent. It contains coarse buckets only — no names, no institutions, no exact counts, no dates finer than a quarter, and no reference to the session it came from. It is anonymous rather than pseudonymous, which is why it can be published openly. Ticking the opt-out box when opening a session removes the option altogether.

No tracking of any kind

No cookies, no analytics, no tracking, no advertising, no third-party scripts, and no accounts. Candidate data is never used to train anything.

Who is responsible

The committee running the selection is the controller of its candidates' data and is responsible for informing them, for its own legal basis, and for retention. AKC supplies the software and operates the open dataset.

Limitations

It does not detect bias it has not been told to look for, including racial, disability and socioeconomic bias.
It cannot verify the data entered. A wrong h-index produces a confident wrong flag.
Gender is recorded as entered by the chair, not as self-identified by the candidate.
Double-blind mode hides names and gender from members, but a career stage, an institution and an h-index may still identify someone in a small field. It reduces bias; it does not guarantee anonymity. The chair is never blind.
It has no memory across sessions, so it cannot yet show your committee's patterns over time.
The chair knows which seat link went to which person, but never learns which ballot came from which seat.

Seat 1