Privacy Policy

R3ciprocity, LLC provides research discovery, private project tools, Community features and free and paid services through its website and mobile application. This policy explains how we collect, use, share and retain information when you use the Service. The information processed depends on the features, payment channel and optional choices you use.

Private projects and unpublished research are private by default and are only shared for feedback when you explicitly enable sharing. Account activity helps personalize paper recommendations. Public scholarly records remain distinct from your own writing and private activity.

Contact contact@r3ciprocity.com for privacy, access, correction, export or deletion questions. Our contact location is Tallahassee, Florida 32311, United States. Do not send passwords, API keys or confidential manuscript text in ordinary email.

USER-CREATED PRIVATE CONTENT includes project titles and drafts, unpublished proposals, saved private writing, project steps and versions, notes, user-supplied contribution evidence, private document originals and any extracted text. Feedback requests can contain a selected excerpt of that work. Storing a draft or attaching a source does not make it public.

PUBLIC SCHOLARLY DATA includes paper titles, authors, affiliations, identifiers such as DOI, ORCID and OpenAlex IDs, source links, publication years, topics, available licence information, citation counts and dated observations. An accessible scholarly record is not proof that the person viewing or linking it is its author. Public availability does not transfer copyright in a paper or abstract.

PLATFORM ACTIVITY DATA includes sign-in/account identifiers, settings, papers served, timestamped citation choices, likes/unlikes, saved reading, skips and explicit rejections, approximate active reading time and scroll depth, comprehension responses, difficulty preferences, project actions, review requests and responses, poll answers, predictions, optional confidence, earned progression, support reports and notification preferences/statuses. These are linked to an account internally; an aggregate display does not make the underlying records anonymous.

AI-PROCESSED CONTENT is a purpose-specific subset of writing, related feedback, selected project steps or earlier writing, and eligible scholarly excerpts submitted through a consented AI tool. Generated suggestions, validation results, cached output and call metadata are also AI-related information. A private project is not automatically an AI submission, and an AI submission is not Community publication.

R3ciprocity uses Amazon Cognito for managed authentication, account verification, recovery and sign-in security. Access controls and multi-factor authentication requirements apply according to the sign-in flow. Cognito processes credentials, email and verification status, MFA information and authentication/recovery requests. R3ciprocity uses the signed account subject to derive an internal owner identifier and separate each person's data. Email verification codes and their verification are handled through Cognito; providing an email address does not itself establish that it is verified.

The browser receives a secure, HTTP-only session cookie and a short-lived login-flow cookie. Provider access/refresh tokens are held by the server, not exposed through browser JavaScript. The current browser session can renew within a bounded session window. Sign-in attempts can produce limited security-stage/error records.

When you choose Google or Facebook sign-in, the selected provider processes authentication and supplies account identifiers and profile information authorized through that sign-in flow. Social sign-in does not give R3ciprocity access to your provider password or replace the Service’s account access controls.

The iOS client uses the system authentication session, PKCE and device-only Keychain storage for native tokens, with account-scoped app caching and logout cleanup.

New project editing starts with feedback sharing off. A saved project must be explicitly marked feedback-eligible to enter the project comparison/game workflow. Separate Community review requests require a chosen sharing mode and explicit confirmation before dispatch to selected reviewers. Private requests are not sent.

A shared project comparison can show its title and short opening excerpt. A matched review request can show the excerpt, question/context and the avatar information needed for that interaction, according to the chosen matched/anonymous mode and identity controls. Anonymous here means identity is withheld from the recipient in that workflow; the service still knows the accounts involved.

Turning off sharing stops new delivery and the application checks current sharing/version status before serving feedback. Recipients may already have read or copied an intentionally shared excerpt or downloaded feedback; withdrawal cannot recall their independent copies. Reviewers must respect the content standards and confidentiality described in the Terms.

General participation and review settings do not by themselves publish every project. Each project/request still needs its explicit sharing choice. Public profile, identity, Portfolio, contribution/service records, Influence evidence, leaderboard membership and comparison-dataset participation have separate controls.

Authorized operation, support, moderation, backup or deletion work may require access to stored records. We do not promise that private data is invisible to the operator or is end-to-end encrypted.

Recommendations use your explicit likes and saved papers, weaker signals from citation choices, approximate reading time/scroll depth and explicit skips/rejections to rank paper topics, fields, journals and text features. Rejecting both papers is a different signal from merely picking one as a citation winner. Comprehension answers and easier/harder preferences help adapt the reading challenge. Time and scrolling are approximate interest clues, not proof of comprehension or research quality.

Reading time is counted when the content is visible, focused and recently interacted with; signals are bounded and normalized. The Service does not record eye movements, a screen video, keystroke recordings or unrelated browsing history. Private project text can also be compared locally with candidate paper content for project relevance.

The preference vector is calculated within the application; this ordinary recommendation path does not send your private vector or project text to an AI service. Turning off reading signals in Profile stops new server-accepted reading signals and their use in that recommendation path. It does not switch off all recommendations, erase existing choices or stop explicit likes/saves from being used. Other controls adjust difficulty and exploration.

R3ciprocity does not make tenure, hiring, promotion, admission, research-quality or publication decisions. Descriptive comparisons and recommendations do not determine those outcomes.

We store review requests, selected excerpts, reviewer assignments, comments/reactions, useful-feedback responses, voluntary interests/availability and related contribution records to run Community matching and feedback. The Community question bank rotates research, PhD life, career, resources and publishing questions. Earlier learning scores are not counted as new poll votes.

Poll choices and original timestamps are saved to your account. With the current aggregate preference enabled, eligible answers can contribute to Community percentages. Turning that preference off excludes your answers from subsequent aggregate reads and hides Community results for your account. It is distinct from optional Google Analytics consent.

Poll percentages require at least five eligible responses. Optional broad, self-declared career-group results require at least ten responses in every nonempty group involved, including any unclassified residual group. Institutions and individual answer records are not displayed in these poll breakdowns. Small samples show that your answer is saved instead of fabricated percentages. The responses are Community opinions, not a representative survey or AI advice.

Review matching, voluntary avatar leaderboards and opted-in profile sharing can disclose limited contribution/context information to other participants. Contributions and moderation are governed by the Terms of Use.

The service stores your selected public paper, answer, prediction type, creation/deadline timestamps, evaluation window, citation baseline, matched-field comparison papers and dated source observations. Portfolio also includes earlier paper selections and their available entry snapshots. Explicit future forecasts are distinguished from historical game choices.

Where enabled, the Forecast / Momentum meter records account-specific progression and unlock receipts, with optional 1x/2x/3x self-confidence. Its confidence labels are not money, spendable bets or extra weight in Community totals.

Community forecast totals require explicit participation and sufficient distinct eligible accounts answering the same question with a comparable frozen baseline and window. The current minimum is ten contributors. Individual account identifiers are withheld from those views. The optional forecast participation choice respects an existing aggregate privacy OFF preference.

Dated citation observations, forecasts, outcomes and crowd-movement snapshots are preserved for later evaluation. Source corrections or withdrawn participation can cause a current display or old crowd point to be withheld rather than rewritten. Preserving dated evidence does not override account erasure. Deleted participation can cause affected crowd history to be removed; other participants’ original predictions and citation outcomes are not recalculated.

These are research-attention forecasts and Community opinions. They do not represent scientific merit or institutional decisions, and provide no gambling, cash settlement, trading or real-money return.

Influence looks up public OpenAlex author/work records, including searches using an ORCID identifier, and saves dated results, caches, report snapshots and selected comparison options to your account. Public author names, affiliations, publication/citation histories, coauthor/citing-paper relationships and topic information can appear in those reports.

The ORCID lookup is an identifier-based OpenAlex lookup. It is not an ORCID OAuth identity verification, does not read restricted ORCID records, and does not edit an ORCID or OpenAlex record. User confirmation links a public identifier to the app account; it does not independently prove authorship or ownership.

User-supplied mentoring, teaching, service, career or contribution evidence is stored separately as self-reported information. Service records are private until you enable the relevant public profile setting and select a record to share. Public-dataset consent is a separate choice; downloading a report does not publish your information in an external comparison dataset.

Public scholarly records can remain in the public source and the platform's shared catalogue when an app account is deleted. App-private notes, the account linkage and your own saved report/selection records are different information subject to applicable account rights.

AI assistance uses OpenAI’s Responses API through the platform’s server-side service. Requests require your confirmation and the relevant source permissions. Private writing and imported source materials are sent only through an enabled AI tool after you confirm that request. Normal game play, Community polls, citation snapshots and local wording cleanup do not require an AI request.

A confirmed writing request may include the draft, related feedback, attached source materials, selected project steps/earlier writing, permitted paper excerpts and bounded topic/context information. Review the proposed inputs and ensure you have authority to submit them, including collaborators' or reviewers' writing. This request does not share the project with other users.

The request sets store:false. That does not establish zero provider retention. OpenAI documents separate API training and retention controls; provider processing depends on the applicable settings and agreements. OpenAI data controls.

The app records the feature, model, request hash, status, timing, available token counts and estimated cost, and can cache validated generated output for 24 hours. Expired cache is removed when the cache-maintenance/request path runs; this is not a guarantee that every copy vanishes exactly at 24 hours. Saved project revisions and suggestions you keep are separate retained records. Do not submit another contributor’s private writing or feedback without authority to have it processed.

We store notification events, categories, read status, preferences, delivery jobs and delivery status to provide and manage account notifications. Optional email and push start off. Sign-in includes a separate, unchecked email opt-in. Declining does not prevent play or affect rewards. After an initial acceptance, reopening the app does not overwrite your saved notification choice; you can change it in Profile → Notifications. Choosing categories/channels affects the product notification queue. Required account/security messages, such as Cognito verification and recovery, are separate.

Optional product messages follow the channels and categories you choose. Registering a device for push notifications requires the device’s permission and a notice explaining the registration information used. You can withdraw optional notification choices in Profile or the relevant device settings.

Account onboarding and support email use managed-authentication or operator-managed email services, including Google Workspace. A support report in the app stores your description, a bounded app path, a broad device category and a timestamp for owner investigation. It does not automatically email the owner. Keep report descriptions free of credentials and unnecessary confidential material.

Optional product emails include a recipient-bound unsubscribe mechanism; it affects the optional product channel, not mandatory account/security processing. Optional product messages follow your saved channel and category preferences.

When you purchase a paid plan or service on the web, Stripe processes the payment details you enter. R3ciprocity stores customer, checkout and subscription identifiers, the selected plan and billing status, and relevant transaction, invoice, payment, refund and billing-notification records. We use these records to provide paid access, manage renewals and cancellations, issue or reconcile refunds, handle billing questions, prevent fraud and meet accounting or legal obligations. R3ciprocity does not collect or store raw card numbers through its application.

Each paid offer states its price, billing interval, renewal, cancellation and refund terms before purchase. Stripe processes payments under its applicable terms and Privacy Policy, including transaction and security information needed for payment processing. For purchases through Apple, Apple handles payment details and supplies the purchase or subscription records needed to provide access. The purchase channel determines how payment details, cancellations and refunds are handled. The Service does not provide connected-account payouts.

The web Service offers opt-in Google Analytics and local optional product-measurement records. The owner is excluded. Consent is required before the Analytics tag is requested; there are no pre-consent analytics pings in the client. Fixed event names and broad screen/placement labels are used instead of project titles, manuscript text, account IDs, paper IDs or full private/query-string URLs. Google still receives the technical request information needed to provide its service, such as network/device information.

Advertising storage, remarketing, Google Signals and ad personalization are disabled in the client. The app has no advertising/data-broker integration. We do not send private writing, account IDs or contact details to Google Analytics through the app’s measurement events.

Turning off optional analytics stops further measurement and clears the app's local identifiable analytics/funnel records and its analytics-cookie controls. It does not erase core game/project data or necessarily delete already transmitted provider data. Google-side erasure/retention must be handled separately.

The live CloudFront access-log delivery is restricted to date, time, HTTP method, response status, edge result type, duration and TLS protocol. Its selected fields exclude client IP, request URI/query, cookies and user agent. Other authentication, application/system and provider security processing can have different records. Operational error codes, job status, support reports, AI usage metadata and deletion/recovery audit entries support reliability and abuse investigation. Security and service-provider records have separate retention; the 28-day analytics-cookie setting does not apply to every infrastructure log.

Optional product analytics are separate from Apple-operated TestFlight diagnostics, described below.

The native client stores account-scoped app state on the device and uses device-only Keychain for native credentials. Browser/app caches can hold paper packs, pending actions, preferences and saved viewing state for fast operation. Logging out ends the active account session. Exported or downloaded files remain on your device until you remove them. Protect your device and avoid leaving an active account on a shared device.

The Service does not access your camera, microphone, contacts, photo library or precise location for these research and project features.

For applications distributed through TestFlight, Apple processes installation, usage/crash diagnostics and submitted feedback, with some information supplied to the developer. Feedback can include screenshots and comments that you choose to submit. Those data are separate from R3ciprocity's opt-in web Analytics. Do not include confidential research in a TestFlight screenshot or comment unnecessarily. Apple TestFlight privacy information.

An application distributed through TestFlight is subject to its published availability and Apple’s applicable terms. Review the privacy information accompanying the installed version and keep private manuscripts out of optional TestFlight feedback.

AWS hosts the Service in the United States, with regional application resources in Ohio and global CloudFront delivery. Cognito handles authentication; EC2/storage, protected S3 backups and operational logging support the app. AWS and authentication providers can process connection/security information to provide their services.

OpenAlex receives public scholarly identifiers/search parameters for public-data retrieval, not your private manuscript through the lookup path. Clicking a paper/ORCID/publisher link opens another site's own data practices. Google Analytics receives consented measurement requests; Google Workspace handles authorized owner-operated emails. OpenAI receives the selected content for AI requests you confirm. Stripe processes web payments and related billing information. Apple processes information associated with its app distribution, device services and purchase channels.

Operator/support personnel can process information to fulfill verified requests and investigate reports. We may disclose relevant information when required by applicable law or valid legal process, or when necessary and lawful to protect rights, safety or the Service. We do not promise disclosure of a private project's identity simply because someone makes an allegation.

International access and service providers can involve processing outside your country, where protections may differ. We use applicable contractual and other safeguards where required by law. Florida jurisdiction in the Terms does not remove privacy rights that applicable law grants you.

We retain private work, account activity, scholarly report snapshots, predictions and support or notification records to provide your account features, maintain saved history, resolve requests and meet legal or security obligations. These records remain until project or account deletion or another applicable retention action. Retention depends on the type of record and its purpose; we do not promise permanent storage.

Profile → Account provides an owner-scoped data export; original uploaded files have separate owner-only downloads. Credentials, session/unsubscribe tokens and counterparty account IDs are excluded from the export. Export a copy before requesting irreversible removal.

Project deletion requires a current confirmation and removes the project's active linked versions, steps, generated drafts, feedback and related project content. Private file cleanup is queued and can remain pending when storage is not configured or cleanup fails. Other projects, earned progression, separately saved public papers, downloaded exports and deliberately copied text are outside that project deletion.

Account deletion confirmation immediately blocks account access and records a durable request. Completion requires verified sign-in/provider cleanup, subscription cancellation and billing reconciliation, private file cleanup, owner-record removal in both app databases and integrity/checkpoint verification. Some deletion steps require operator follow-up, and a request remains pending while required cleanup is unfinished. Deletion is not reported complete until the required cleanup and integrity checks succeed.

Separate backups retain earlier copies. Earlier backup copies remain until removed or replaced through the backup-retention process; deleting an active account does not instantly erase every backup copy. A durable account-deletion journal supports blocking/replaying requests after restore, but is itself a retained protective record. Project deletion and restored backups require operator checks to avoid returning deleted private work to active use. Contact us to follow up on a pending deletion.

Analytics cookies expire 28 days after creation and are not extended on each activity. Local optional event and round records have a rolling 28-day retention, enforced by periodic maintenance; the preference recording your ON/OFF choice is retained until you change it or delete your account. This is distinct from Google's retention and from core activity. Confirmation tokens have short expiries; AI-output cache has the separate expiry described above.

Downloaded copies, recipients' independent copies and public records controlled by other services are not erased by deleting an app account. Information needed for a legal obligation, a pending dispute or security investigation may require separate restricted retention. We explain an applicable exception in response to a verified deletion request.

The Service is offered to users who are 13 years of age or older and reside in the United States or its territories or possessions. You may play a full 30-choice Abstract Game round without signing in. After the last choice, sign in, affirm that you are at least 13 and agree to the Terms of Use to save the round and access your streak file, saved progress and other account features. These required choices are separate from optional email and analytics choices. We record the accepted Terms version, affirmation and acceptance time with your verified account. We do not request a birth date through this affirmation.

During guest play, the game serves eligible public abstracts without issuing account-specific decisions or storing your activity. Your choices and provisional round rewards stay only in the page’s memory: no account record, cookie, local/session storage, IndexedDB, reading telemetry or analytics tag is created for guest play. Refreshing or leaving discards the unsaved round. If you explicitly choose to save after completing 30 choices, the completed choices and their dated citation snapshots are submitted only after verified sign-in, age confirmation and Terms acceptance. Cancelling sign-in leaves the round unsaved in the original open game tab. Loading a page necessarily involves connection information and essential infrastructure/security processing; unsaved play is not a promise that no technical information reaches any server.

We do not knowingly collect personal information from children under 13 through account or recorded play and do not offer a parental-consent account flow. Do not create or use an account if you are under 13. We do not use games or rewards to encourage disclosure of unnecessary personal information. If we learn that a child under 13 has provided personal information without appropriate authorization, we will restrict access and take steps to delete the information, subject to applicable law. Contact contact@r3ciprocity.com to report such information.

Use Profile for sharing, reading, notification, public-profile and optional measurement choices. You may request access, correction, export, deletion or withdrawal of optional consent through contact@r3ciprocity.com. We verify account ownership before disclosing or removing private data. We do not require you to send identity documents in ordinary email. We will respond within applicable legal deadlines, explain any lawful limitation or extension, and provide a way to follow up.

Rights to object, restrict processing, obtain portability or complain to a regulator depend on the law that applies to you. Correcting an account link does not change a public ORCID/OpenAlex record; we can help distinguish an account correction from a source correction. Withdrawing an optional permission does not delete other records needed to provide the Service.

We may update this policy by posting an updated version and date. We will provide additional notice or obtain new consent when required for a material change. A change does not retroactively authorize a new use of private writing or replace your separate optional choices.