Last Updated: November 2025
Contact: info@ihtimam.org
1. Statement of Intent
At İhtimam, we place the ethics of care at the heart of how we think about transforming organizations, communities, and systems. We recognize the possibilities that artificial intelligence technologies offer in terms of speed, accessibility, creativity, and efficiency. At the same time, we position technology not as an end in itself, but as one of the tools that can support people, relationships, and our collective capacity to care.
For us, the ethical use of AI is not limited to whether a system produces accurate results, protects personal data, or avoids discrimination. When we use AI, we also ask:
Whose needs are we responding to? Whose voices are we hearing, and whose might we be missing? Are we separating people from the relationships and circumstances that shape their lives? Is technology creating distance between us and the people affected by our decisions? Whose vulnerability might it increase, and whose capacity might it strengthen? Whose labor does it make visible, and whose does it render invisible?
The ethics of care reminds us that people are not isolated, independent individuals. We live in relationships of dependence and interdependence with one another, other living beings, our communities, and the systems around us.
For this reason, we use AI not to replace human judgment, responsibility, or relationships, but to create more space for them.
2. Our Care-Based Principles for AI
A. Attentiveness — Seeing the Need
We recognize that not every technological possibility needs to become a technological solution.
Before using AI in a process, we ask: “What is actually needed here, and is AI an appropriate tool for meeting that need?”
We do not use AI simply because it is available, faster, or capable of producing more output. Sometimes speaking with another person, reading a text ourselves, thinking together, or using no technological tool at all is the better option.
We take into account that AI systems can make mistakes, generate false information, reproduce existing social biases, and simplify complex human experiences.
In practice: Before starting a task, we ask not only “Can we do this with AI?” but also “Should we do this with AI?”
B. Relationality and Interdependence — Keeping People in Context
From an ethics of care perspective, we do not see people as collections of individual attributes or data points. We recognize that all of us live within relationships, responsibilities, dependencies, and interdependencies.
We know that AI can make these relationships invisible when it classifies, summarizes, or makes inferences about people.
For this reason, particularly when AI outputs concern people or have consequences for them, we do not reduce people to a category, score, profile, or “average user.”
We take into account that a person’s caring responsibilities, access to resources, relationships, and experiences of inequality can fundamentally change the meaning of a situation.
In practice: Alongside the patterns AI shows us, we actively look for those who fall outside the pattern.
C. Proximity — Not Increasing Moral Distance
We recognize that AI can increase the distance between those making decisions and those affected by them. When direct encounters with people are replaced by datasets, categories, scores, and automated processes, it can become harder to see the real consequences of our decisions.
We therefore use AI not to see people less, but to understand their circumstances and needs more fully.
We do not allow automation to obscure responsibility or become an excuse to say, “the system said so.”
In particular, we do not use AI outputs as the sole decision-maker in processes that affect people’s access to opportunities, resources, programs, or communities.
In practice: When an AI output affects a person, we ask:
“If this person were sitting in front of us, what context would we want to know that AI might currently be hiding from view?”
D. Responsibility and Human Agency — Not Delegating Judgment to Machines
We do not regard AI as a moral agent. When we choose to use an AI-generated output, we take responsibility for how it is used and for the effects that follow.
“AI said so,” “the algorithm selected it,” or “the system generated it” are not explanations that remove responsibility.
We do not delegate strategic, ethical, emotional, or otherwise consequential decisions about people entirely to AI.
Nor do we consider it sufficient for a person to simply see an output at the end of an automated process. For human oversight to be meaningful, people need to be able to question, modify, and, when necessary, reject an AI output altogether.
In practice: For significant AI-assisted outputs, there is a human being who holds final responsibility.
E. Context and Plurality — Seeing Beyond the Pattern
AI is powerful at identifying patterns across large amounts of information. The ethics of care, however, reminds us that what is common is not always the same as what is important.
An experience expressed by only a small number of people in a research process is not insignificant simply because it is uncommon. A less common experience may reveal a need, inequality, or vulnerability that existing systems fail to see.
For this reason, when we use AI to support research, summarization, or synthesis, we look not only for dominant themes but also for experiences that are underrepresented, contradictory, exceptional, or easily made invisible.
We do not reduce diverse human experiences to a single “average” experience.
In practice: In AI-assisted synthesis, we also ask:
“Who or what is missing from this generalization?”
F. Attending to Vulnerability — Not Penalizing Vulnerability
We understand vulnerability not as an exceptional state, but as part of the shared condition of human life.
We recognize that AI systems can reproduce existing social inequalities and can misread the experiences of people who are underrepresented, have fewer resources, or experience discrimination.
We ensure that a person’s vulnerability does not simply become a variable working against them. We do not treat caring responsibilities, disability, economic circumstances, language, geography, or similar contextual factors solely as deficits or “risks.”
We consider who an AI system does not work well for to be just as important as who it works well for.
G. Voice, Participation, and Contestability — Giving People a Say in What Is Said About Them
We do not see people simply as subjects from whom data is collected, who are analyzed, or about whom decisions are made. We recognize people as holders of knowledge about their own experiences.
For this reason, wherever possible, we enable people affected by AI-supported processes to have a say in how their experiences are represented.
When an AI system misclassifies people, misrepresents their experiences, or overlooks important context, we make it possible for those outputs to be questioned, corrected, and, where necessary, rejected.
Particularly in research, evaluation, and participatory processes, we recognize that an AI system’s interpretation of what someone has said is not the same thing as what that person actually said.
The fact that someone has shared an experience with us does not mean they have given us permission to generate unlimited new inferences about that experience through AI.
H. Competence — Critical AI Literacy
We do not understand responsible AI use simply as knowing how to write good prompts.
We see AI literacy as understanding the possibilities and limitations of AI, its modes of failure, data security risks, social biases, and ethical consequences.
We treat AI-generated texts, analyses, and images as outputs to be evaluated rather than finished products.
We verify significant information generated by AI against reliable sources and take responsibility for content published in İhtimam’s name.
I. Trust, Privacy, and Transparency
We do not see information shared with us by our stakeholders simply as a “data source.” We see it as part of the relationship between us.
We do not enter personal, sensitive, or confidential information into publicly available AI systems that lack appropriate safeguards.
For us, privacy means more than removing someone’s name from a dataset. We recognize that generating sensitive, unexpected, or undisclosed inferences about people from anonymized information can also have ethical consequences.
When AI has played a meaningful role in producing a piece of research, a report, a strategy, or a creative work, we disclose this appropriately. We distinguish between minor uses, such as spelling or language checks, and uses that substantially shape the content itself.
We understand transparency not simply as saying “we used AI,” but, where relevant, as making clear where we used it and for what purpose.
J. Caring for Labor — Not Making Invisible Labor Even More Invisible
We do not see AI systems simply as technological products. We recognize the visible and invisible human labor behind their development and use, including data production, data labeling, content moderation, creative work, maintenance, and other forms of labor.
In our own work, we ensure that the efficiencies created by AI do not erase the human labor that makes our work possible.
We acknowledge the contributions of participants, facilitators, caregivers, researchers, translators, designers, and others whose often less visible labor makes a report, research process, event, or other piece of work possible.
We do not automatically turn time saved through AI into expectations of greater productivity. We also use the capacity opened up by technology for thinking, learning, resting, building relationships, and caring for one another.
3. How We Use AI
We use AI in areas where it supports human capacity and relationships.
Idea development: We use it to explore different perspectives, generate questions, and move through creative blocks.
Research, summarization, and synthesis: We use it to identify patterns across large amounts of information and create initial syntheses. We review these outputs ourselves and look not only for dominant patterns, but also for contradictions, exceptions, and less-heard voices.
Language and editing: We use it to improve clarity and accessibility, support translation, and assist with language review.
Visualization: We use it to make concepts visible, experiment with ideas, and develop conceptual designs and visual drafts.
Repetitive tasks: We use it for appropriate administrative and repetitive processes in order to free human time for work that requires relationships, reflection, and care.
We do not treat AI-generated inferences about a person’s or group’s life, needs, intentions, emotions, or vulnerabilities as reality itself.
Where We Do Not Use AI, or Where We Restrict Its Use
We do not process sensitive personal or organizational information in systems without appropriate safeguards.
We do not rely solely on AI to assess a person’s emotional state, character, intentions, or trustworthiness.
We do not delegate final decisions determining people’s access to opportunities, resources, programs, or communities to AI.
We do not substitute AI for human relationships in situations involving crisis, conflict, grief, sensitive feedback, one-to-one support, or deep listening.
We do not use personal experiences shared with us for a specific purpose to conduct new AI analyses or profiling that the people concerned could not reasonably have anticipated.
We do not choose tools or forms of use that clearly violate copyright or creative labor rights. Where there is uncertainty about the sources of a tool’s training data, we include that uncertainty in our ethical assessment.
4. Questions We Ask When Using AI
Not every use of AI requires the same level of ethical scrutiny. As the potential impact on people, relationships, or resources increases, we pay closer attention to the following questions:
- Need: Do we actually need AI here?
- Relationship: Does this use strengthen relationships between people, or introduce new distance?
- Context: What circumstances that AI cannot see might be important for understanding this situation?
- Voice: Who becomes visible in the output, and who becomes invisible? Who falls outside the pattern?
- Vulnerability: Could this use increase the vulnerability of someone who already holds less power?
- Power: What is the power relationship between the person using AI and the person affected by its output? Can the affected person speak, add context, or contest the outcome?
- Privacy and reasonable expectations: Would people reasonably expect the information they provided to be used in this way?
- Responsibility: If something goes wrong, who will notice it, correct it, and take responsibility?
- Labor: Whose labor does this use reduce, increase, or make invisible?
- Alternative: Could we meet the same need through a simpler, less resource-intensive, or more relational approach?
We do not use these questions as an “ethical compliance test.” We use them as a practice of care that helps us pause, direct our attention, and look again.
5. Caring for the Planet and the Material Conditions of Technology
We recognize that artificial intelligence is not an abstract process that simply happens “in the cloud.”
AI infrastructure depends on material systems involving data centers, energy, water, hardware, minerals, and human labor.
For this reason, we do not use AI simply because we can. When a simpler method can meet the same need, we choose it. When we do use AI, we use a model proportionate to the task rather than treating greater computational capacity as inherently better.
We do not see technological efficiency as a justification for unlimited use.
For us, care in technology extends beyond the user. It also entails responsibility toward the people, labor, other living beings, and the planet that make technology possible.
6. Feedback, Repair, and Learning
We understand the ethics of care not as a set of perfect rules, but as an ongoing practice that requires us to pay attention, take responsibility, respond competently, listen, and change our behavior when needed.
This policy is therefore not a static document.
As AI technologies evolve, our uses of them change, and our understanding of the ethics of care deepens, we review and update this policy.
When we recognize that our use of AI has misrepresented, rendered invisible, excluded, or harmed a person or community, we do not stop at acknowledging what happened. Where possible, we correct, learn, and repair.
We recognize that there may still be things we have not yet seen.
We welcome you telling us: “There may be a more caring way to do this.”
We see feedback not simply as criticism, but as a contribution to the development of our practice of care. We respond within 48 hours on working days.
Intellectual Grounding
This policy draws on İhtimam’s approach to the ethics of care as well as scholarship that examines AI ethics through relationships, interdependence, context, vulnerability, and power.
In particular, Carolina Villegas-Galaviz and Kirsten Martin’s work on moral distance draws attention to the ways AI can increase the distance between those making decisions and the people affected by them. Their approach invites us to evaluate AI ethics not only through technical performance, general principles, or individual rights, but also through people’s relationships, particular circumstances, vulnerabilities, and the inclusion of different voices.
At İhtimam, we bring this perspective into conversation with our own practice of the ethics of care.
For us, care-centered AI use is not only about preventing harm. We also look at whether technology distances people from one another and from the consequences of their decisions; whether it strengthens human agency, plurality, and relationships; how it affects invisible labor; and whether it creates more room for care.
We use AI not to know less about one another, but to create more room to care for one another.