Ethical Data Handling for AI and Information Systems Research

Artificial intelligence and information systems research increasingly depend on large, complex, and sometimes highly sensitive datasets. Training data, user interactions, behavioral records, sensor streams, medical or financial information, system logs, images, text, and publicly collected online data can all contribute to the development and evaluation of modern technologies. Yet access to data does not automatically mean that it is ethically appropriate to collect, process, share, or reuse.

For researchers, ethical data handling is therefore not simply a matter of complying with institutional requirements. It is part of producing research that is credible, reproducible, responsible, and worthy of public trust.

This issue is particularly relevant to the Ubiquitous Technology Journal (UTJ), published by Crosslink Studies (CLS). UTJ covers computer science and engineering research across areas including artificial intelligence, IoT, human-computer interaction, cybersecurity, ubiquitous computing, machine learning, smart environments, and privacy-preserving technologies. Its scope explicitly includes ethical considerations of AI and data privacy in ubiquitous environments.

What Does Ethical Data Handling Mean?

Ethical data handling refers to the responsible collection, storage, analysis, sharing, and disposal of research data. The central question is not simply “Can this data be obtained?” but rather:

“Should this data be collected and used in this way, and can its use be justified to the people and communities represented by it?”

A responsible approach considers privacy, informed consent, security, fairness, transparency, data minimization, ownership, and the potential consequences of misuse.

1. Collect Only the Data You Actually Need

Ethical research begins at the collection stage. Researchers should define what information is necessary to answer the research question and avoid collecting unrelated personal or sensitive information simply because it is technically available. This principle of data minimization reduces privacy risks and makes subsequent data management more straightforward.

Before collecting information, researchers should ask:

• What data is necessary?
• Why is each variable required?
• Who could be affected?
• How long will the data be retained?
• Could the same research objective be achieved with less sensitive information?

2. Obtain Appropriate Consent and Ethical Approval

When research involves human participants or potentially identifiable information, researchers must carefully consider applicable ethical requirements. Consent should be meaningful and understandable. Participants should know, where applicable, what information is being collected, why it is being used, how it will be protected, and whether it may be shared for future research. Ethical review may also be required depending on the research design, institution, jurisdiction, and type of data involved.

Researchers should therefore distinguish between:

Anonymization — information is processed so that individuals cannot reasonably be identified.

Pseudonymization — identifying information is replaced with codes or pseudonyms, but re-identification may still be possible using additional information.

3. Examine Bias Before Training or Analysis

Ethical data handling also requires researchers to consider what the dataset represents—and what it does not represent. AI systems can reproduce or amplify biases present in training data. A dataset may underrepresent particular populations, contain historical inequalities, or reflect patterns that are inappropriate for automated decision-making.

Researchers should therefore examine:

• Sampling procedures
• Missing groups or populations
• Class imbalance
• Label quality
• Historical bias
• Data collection conditions
• Potential discriminatory outcomes

4. Document Data Sources and Processing Decisions

Transparency is essential for reproducible research. Researchers should document where datasets originated, how they were collected, what preprocessing was performed, which observations were excluded, and how variables were transformed. For AI research, useful documentation may include dataset source, collection period, inclusion and exclusion criteria, preprocessing procedures and labeling methodology.

5. Handle Publicly Available Data Responsibly

One of the most common misconceptions in digital research is that information available online is automatically free for unrestricted research use. Public accessibility does not eliminate ethical responsibilities. Researchers collecting information from websites, social platforms, repositories, or other online sources should consider the platform’s terms, applicable laws, privacy expectations, copyright, and the sensitivity of the information.

6. Use AI Tools Without Compromising Confidential Data

The growing availability of generative AI tools introduces another important dimension of research ethics. Researchers may use AI systems for coding assistance, language editing, data processing, or analytical support. However, confidential datasets, unpublished manuscripts, identifiable participant information, proprietary code, or restricted research materials should not be entered into external AI systems without first assessing the relevant privacy, security, ownership, and data-use terms.

7. Share Data Responsibly Not Automatically

Open research and data sharing can improve transparency and reproducibility, but ethical data sharing does not mean publishing every dataset without restriction. When appropriate, researchers can use trusted repositories, De-identified datasets, controlled-access repositories and data-use agreements.

8. Report Limitations Honestly

Ethical research requires researchers to communicate limitations transparently. If a dataset is small, geographically restricted, biased toward a particular population, incomplete, obtained under license, or subject to access restrictions, those limitations should be acknowledged. Similarly, researchers should not selectively report results simply because they support the proposed hypothesis.

A Practical Ethical Data Checklist

Before submitting an AI or Information Systems manuscript, researchers should ask:

Data Collection

• Was the data collected for a legitimate research purpose?
• Was appropriate consent or authorization obtained?
• Was unnecessary personal information avoided?

Privacy and Security

• Have identifiers been removed or protected?
• Is sensitive data securely stored?
• Are access permissions appropriate?

Analysis

• Have potential biases been examined?
• Are preprocessing and exclusion decisions documented?
• Are results reported accurately?

AI Use

• Were confidential materials protected when using AI tools?
• Is AI use disclosed where required?
• Were AI-generated outputs independently checked?

Sharing

• Can supporting data or code be responsibly shared?
• Is a trusted repository available?
• If sharing is restricted, is the reason clearly documented?

Ethical Data Handling as Part of Research Quality

Ethical data practices should not be separated from scientific quality. The way data is collected and managed can directly affect the validity, fairness, reproducibility, and interpretation of research findings. This is particularly important for the technology disciplines represented by UTJ, where AI systems, connected devices, information platforms, and data-driven applications increasingly influence real-world decisions. UTJ’s published work already includes research examining ethical AI design and responsible innovation, demonstrating the relevance of these questions to the journal’s broader research community.

For CLS authors, responsible data handling should therefore be considered throughout the research lifecycle: design, collection, storage, analysis, sharing, and publication.