DMP Guide: NSF for Computer Science & AI

1. Funder Policy & Open Data Compliance

In alignment with international open-science mandates, National Science Foundation requires all principal investigators to submit a comprehensive Data Management Plan (DMP) with their grant application. A robust DMP details how research data will be collected, processed, documented, stored, shared, and preserved both during and after the project.

For projects in the field of Computer Science & AI, managing data correctly is essential not only for compliance, but also to support peer-review validation and reproducibility. All DMPs must be submitted through the Research.gov portal, using standard institutional guidelines.

2. Data Types, Formats, and Metadata Standards

A high-quality DMP must explicitly identify the types of data that will be generated and specify open, non-proprietary file formats to ensure long-term usability. For Computer Science & AI, datasets typically range from raw observational measurements to curated computational models.

Simulations and code-based datasets for **Computer Science & AI** must be packaged alongside the exact processing scripts and execution environments. Documenting this containerized configuration guarantees that **NSF** audits can reconstruct each analysis step.

To guarantee discoverability, datasets should be documented using standardised metadata schemas that map to the Mathematical Concepts branch of scholarly vocabularies. This ensures indexers and crawlers can crawl and identify research outputs accurately.

3. Step-by-Step DMP Construction Protocol

When preparing your DMP for a NSF proposal, structure your document around these core sections:

1. Data Collection and Generation:
   Describe the methodology, instrumentation, or software used to collect or generate new data. Detail quality assurance and quality control measures implemented at your facility.
2. Documentation and Metadata:
   Explain how the data will be documented, including accompanying read-me files, data dictionaries, and laboratory notebooks. Specify the metadata standards to be utilized (using CodeMeta schema, Schema.org definitions, Software Ontology (SWO) as standard).
3. Ethics, Intellectual Property, and Consent:
   Address how sensitive or confidential datasets will be handled. Detail anonymisation processes, access controls, and compliance with institutional ethics boards.
4. Storage, Backups, and Security:
   State where data will be stored during active research. Detail automated backup schedules, server redundancies, and access authorisation protocols.
5. Long-Term Preservation and Archiving:
   Select the digital repository for post-project archiving (such as Zenodo (integrated with GitHub), Software Heritage, Figshare, and directory servers mapped in IEEE Xplore, ACM Digital Library & arXiv). Confirm that the repository supports persistent identifiers (handles/DOIs) and provides secure preservation.

Open Science Workflows, Data Curation & Repositories

Establishing a robust data management plan dmp for Computer Science & AI requires outlining rigorous data collection methods alongside established data curation standards from day one. PIs can leverage structured dmptool workflows to coordinate these data frameworks for review by National Science Foundation. Our project methodology mandates systematic data cleaning and continuous verification of data integrity to support reproducible data wrangling pipelines. To aid secondary usage, a comprehensive data dictionary will accompany every published record. For active storage, the proposal compares a relational data warehouse schema against an unstructured data lake model, reviewing the functional benefits of a data lake vs data warehouse environment for general data analysis and initial exploratory data analysis of study outputs. Open-access guidelines require teams to push finalised files to the dryad data repository, configure shared figshare datasets, or initiate a zenodo data upload, securing a permanent slot in the data citation index that aligns with nsf data management plan directives. To support replication, we will establish strict data versioning protocols on the open science framework osf to guide reproducible data sharing that follows fair data principles examples. When working with sensitive community records, the project will strictly observe the care data principles and indigenous data sovereignty care guidelines to guarantee ethical data stewardship in accordance with NSF rules. Implementing this storage layout satisfies compliance protocols overseen by the NSF data audit team.

4. Frequently Asked Questions