Interpersonal, Emotional and Inclusive Communication

Introduction

Communication in artificial intelligence and machine learning is not only the exchange of technical information. It is also the management of relationships, emotions, identities, expectations and power. A technically accurate message can fail if it is delivered without empathy, expressed ambiguously, or interpreted as dismissive. Conversely, respectful and inclusive communication improves collaboration, psychological safety, requirements gathering, conflict resolution and ethical decision-making.

AI and ML students commonly work in diverse teams involving programmers, domain experts, managers, users, policy specialists and affected communities. These participants may differ in culture, language, gender, disability, professional status, age, socioeconomic background and familiarity with technology. Interpersonal, emotional and inclusive communication enables such teams to collaborate without allowing stereotypes, exclusion or unequal power to undermine outcomes.

This chapter develops four connected capabilities: empathy, assertiveness, self-awareness and relationship management. It then applies them to inclusive communication across differences in ability, identity, culture and power. The chapter also examines discriminatory language, workplace conflict, digital communication and the ethical responsibilities of AI professionals.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain the role of interpersonal communication in academic and professional collaboration.

2. Describe emotional intelligence and apply its main dimensions to communication situations.

3. Distinguish assertive, passive, aggressive and passive-aggressive communication.

4. Use active listening, feedback and conflict-management techniques.

5. Analyse how culture, identity, ability and power influence communication.

6. Identify stereotypes, microaggressions and discriminatory language.

7. Design more inclusive communication practices for AI and ML teams.

8. Evaluate the ethical and practical consequences of communication choices.

1. Interpersonal Communication as Relational Practice

Interpersonal communication is the process through which people create shared meaning in interaction. It includes spoken language, writing, listening, facial expression, gesture, timing, silence and the interpretation of context. It is relational because every interaction communicates not only information but also attitudes about respect, status, trust and belonging.

A useful way to analyse a message is to separate four elements:

• Content: What information is being communicated?
• Intent: What does the sender want to achieve?
• Impact: How does the message affect or appear to affect the receiver?
• Context: What social, cultural, organisational or technological conditions surround it?

These elements do not always align. A team leader may intend to motivate a student by saying, “You need to be more realistic,” but the impact may be discouragement or humiliation. Effective communicators therefore examine impact rather than relying only on intention.

Communication is also transactional. Participants simultaneously send, receive and interpret messages. Meaning is negotiated rather than transmitted perfectly from one mind to another. Noise may be physical, technological, linguistic, emotional or semantic. For example, a video-call delay can disrupt turn-taking; unfamiliar technical terms can exclude a non-specialist; anxiety can cause neutral feedback to be interpreted as hostility.

Verbal and Non-verbal Communication

Verbal communication includes word choice, sentence structure, tone and degree of directness. Non-verbal communication includes posture, eye contact, vocal pace, volume, facial expression and physical distance. In online environments, formatting, response time, punctuation, emojis and camera use may perform similar relational functions.

Non-verbal conventions vary across cultures and individuals. Eye contact may indicate attention in one context but disrespect or discomfort in another. A pause may signal thoughtfulness rather than uncertainty. Effective communication avoids treating one cultural style as universally correct.

2. Emotional Intelligence and Self-Awareness

Emotional intelligence refers broadly to the ability to perceive, understand, use and regulate emotions in oneself and in social interaction. Popular accounts often identify self-awareness, self-management, social awareness and relationship management as major dimensions. These dimensions are useful as a practical framework, although emotional intelligence should not be treated as a fixed personal trait or as a justification for judging people’s emotions.

Self-awareness

Self-awareness is the ability to recognise one’s emotions, assumptions, values, communication habits and effects on others. A student who notices defensiveness during code review can pause before replying. A manager who recognises a preference for rapid speakers can deliberately invite contributions from reflective colleagues.

Self-awareness includes awareness of positionality: the social and organisational position from which one communicates. A senior engineer, for example, may believe that a discussion is open while junior members perceive substantial risk in disagreeing. Awareness of power helps explain why the same words may carry different consequences depending on who says them.

Useful self-awareness practices include maintaining a reflective journal, requesting specific feedback, reviewing difficult interactions and identifying recurring triggers. Reflection should be evidence-based rather than self-condemning. The purpose is improvement, not guilt.

Self-management

Self-management involves regulating reactions so that emotions inform rather than control behaviour. Regulation does not mean suppressing emotion or pretending that injustice is acceptable. It may involve taking a pause, asking for clarification, postponing a response or choosing a more constructive channel.

A practical sequence is:

1. Notice the emotional response.

2. Name it accurately, such as frustration, embarrassment or anxiety.

3. Identify the interpretation producing the response.

4. Check whether alternative interpretations are plausible.

5. Select a response consistent with professional and ethical goals.

This process is especially valuable in online communication, where immediate replies can create durable records and where missing non-verbal cues increase misunderstanding.

Empathy and perspective-taking

Empathy is the effort to understand another person’s experience and communicate that understanding. Cognitive empathy concerns understanding another perspective; emotional empathy concerns resonating with another’s feelings; compassionate empathy adds a willingness to respond helpfully. Empathy does not require agreement, identical experience or abandonment of standards.

An empathetic response might be: “I understand that the deadline and unclear requirements created pressure. Let us clarify the deliverables and agree on a realistic plan.” This acknowledges experience while moving toward action. Empathy must not become performative language without practical support.

3. Assertiveness and Respectful Expression

Assertiveness is the direct, honest and respectful expression of one’s ideas, needs, boundaries and rights while recognising those of others. It differs from aggression, which seeks to dominate or harm, and from passivity, which avoids expressing legitimate needs. Passive-aggressive communication expresses resistance indirectly through sarcasm, delay or withdrawal.

| Style | Typical pattern | Likely effect |

|---|---|---|

| Passive | Avoids disagreement and suppresses needs | Unclear expectations and accumulated resentment |

| Aggressive | Uses threat, insult or domination | Fear, defensiveness and damaged trust |

| Passive-aggressive | Communicates hostility indirectly | Confusion and unresolved conflict |

| Assertive | Clear, respectful and evidence-based | Better coordination and mutual understanding |

An assertive message can use the following structure:

• Describe the observable situation.
• State its effect or significance.
• Express a clear need or request.
• Invite coordination where appropriate.

For example: “The dataset definition changed after the model evaluation began. This makes the results difficult to compare. I need us to document the version used and agree on one evaluation set before continuing. Can we do that in today’s meeting?”

This approach is stronger than “You always change requirements,” because it focuses on observable behaviour rather than character judgement. It also avoids false politeness that hides an important concern.

Boundaries and refusal

Professional assertiveness includes saying no. A responsible refusal identifies the constraint and, where possible, proposes an alternative: “I cannot complete a reliable bias analysis by tomorrow because the subgroup data are not available. I can provide a preliminary risk review tomorrow and a fuller analysis next week.” Such communication protects quality and prevents unrealistic commitments.

4. Active Listening, Feedback and Relationship Management

Active listening

Active listening is a deliberate process of attending, interpreting, checking understanding and responding. It includes:

• giving attention without preparing a premature reply;
• paraphrasing key points;
• asking open and clarifying questions;
• acknowledging emotion or concern;
• summarising decisions and unresolved issues.

A useful distinction is between hearing and listening. Hearing is sensory reception; listening involves cognitive and relational engagement. In meetings, a listener might say, “If I understand correctly, your concern is not that automation is impossible, but that the current data do not represent rural users. Is that accurate?” This both checks meaning and signals respect.

Listening should not be confused with passive agreement. One can listen carefully and still challenge evidence or reject a proposal.

Constructive feedback

Effective feedback is specific, timely, behaviour-focused and oriented toward improvement. The popular “feedback sandwich” is not always effective because praise may appear manipulative or obscure the main issue. A clearer model is:

1. Identify the context or behaviour.

2. Explain its impact using evidence.

3. State the desired change.

4. Offer support or invite discussion.

Example: “In the presentation, the accuracy claim was made without identifying the test population. The audience therefore could not judge generalisability. In the revision, please report the population and limitations; I can help locate the evaluation details.”

Feedback should also flow upward and across teams. A culture in which only junior staff receive feedback is not genuinely developmental.

Relationship management

Relationship management involves building trust, coordinating expectations, handling disagreement and maintaining professional connections. Trust develops through competence, reliability, integrity and care. It is damaged by broken commitments, hidden decisions, ridicule and inconsistent treatment.

Relationships are strengthened by clarifying roles, documenting decisions, acknowledging contributions, sharing credit and repairing harm. An apology is most effective when it identifies the behaviour, recognises impact, avoids excuses and specifies corrective action: “I interrupted you repeatedly in the meeting. That prevented your explanation from being heard. I apologise, and I will let you finish before responding.”

5. Inclusion Across Difference

Inclusive communication aims to ensure that people can participate, be respected and influence decisions. It does not require identical treatment in every circumstance. Fairness may require reasonable adjustments, accessible materials, flexible participation or additional explanation.

Ability and accessibility

Disability-inclusive communication avoids assumptions about what people can or cannot do. Use accessible documents with heading structures, meaningful alternative text, readable contrast and captions where appropriate. In meetings, distribute materials in advance, describe visual information, allow multiple ways to contribute and avoid treating speech speed or eye contact as measures of intelligence.

Use person-first or identity-first language according to individual or community preference. Do not describe accessibility as an inconvenience or portray disabled colleagues as inspirational merely for participating. Ask, respectfully, “What communication or access arrangements would support your participation?” rather than making assumptions.

Accessibility is particularly relevant to AI products. A voice interface, biometric system or automated decision tool may exclude users if design teams fail to consider sensory, cognitive, motor and linguistic diversity.

Gender, sexuality and identity

Inclusive language avoids unnecessary gendering and respects names, pronouns and self-identification. Introduce oneself with a name and, when appropriate, pronouns without pressuring others to disclose. Correct mistakes briefly and continue rather than demanding emotional labour from the person affected.

Avoid assumptions such as “the technical person is probably a man” or “women are naturally better at documentation.” These stereotypes allocate opportunities unequally and distort recruitment, evaluation and collaboration.

Culture and language

Culture influences conventions concerning directness, hierarchy, silence, disagreement, time and politeness. However, cultural explanations should not become stereotypes. Individuals are not representatives of a single fixed culture.

In multilingual teams, use plain language, avoid idioms and explain acronyms. Allow processing time and do not equate accent with competence. Confirm critical decisions in writing. Instead of asking a colleague to “speak properly,” identify the actual communication need, such as volume or audio quality.

Power and status

Power may arise from role, expertise, seniority, employment security, access to information, language fluency or control over resources. Power affects who speaks, whose interpretation is believed and who can safely challenge a decision.

Inclusive meetings can counter these effects by sharing agendas, rotating facilitation, inviting views before dominant voices set the direction, recording dissenting evidence and providing private channels for concerns. Leaders must distinguish consultation from participation: asking for opinions without allowing them to affect decisions can create the appearance of inclusion without its substance.

6. Stereotypes, Microaggressions and Discriminatory Language

A stereotype is a simplified and generalised belief about a group. Prejudice is an evaluative attitude, while discrimination is unequal treatment or outcomes. These concepts may operate explicitly or implicitly. In technical organisations, bias can appear in task allocation, performance descriptions, interruptions, humour, hiring language and assumptions about expertise.

A microaggression is a commonplace verbal, behavioural or environmental message that communicates derogatory or exclusionary meanings toward a marginalised group, often without conscious intent. Examples include repeatedly mispronouncing a person’s name after correction, expressing surprise that someone “understands advanced mathematics,” or asking a colleague to speak for an entire group.

Responses should be proportionate and context-sensitive. A bystander can:

• interrupt: “Let us avoid generalising about that group.”
• clarify: “What evidence supports that assumption?”
• redirect: “Please address the person’s proposal rather than their identity.”
• support privately: “I noticed what happened. How would you like me to help?”
• report persistent or serious conduct through appropriate procedures.

Language concerning race, caste, ethnicity, religion, disability, gender, sexuality, age or nationality requires particular care. Do not repeat slurs merely to discuss them unless an educational or investigative purpose clearly justifies it. Policies should define prohibited conduct, reporting routes, confidentiality limits and protections against retaliation.

7. Conflict Management and Difficult Conversations

Conflict is not necessarily harmful. Task conflict can reveal risks and improve decisions; relationship conflict involving personal attacks usually damages collaboration. The aim is not to eliminate disagreement but to manage it fairly and productively.

A structured conversation may follow these stages:

1. Prepare by identifying the issue, evidence, desired outcome and emotional state.

2. Choose an appropriate private or public setting.

3. Describe observations without exaggeration.

4. Explain impact and relevant values.

5. Listen to the other account.

6. Identify shared interests.

7. Agree on specific actions, responsibilities and follow-up.

For example, a disagreement about a model deployment should distinguish technical evidence from status claims. “The validation set does not include the target age group” is more useful than “You do not understand evaluation.”

Some situations should not be handled through informal dialogue alone. Threats, harassment, discrimination, retaliation, serious power imbalances and safety risks require institutional procedures. Mediation is inappropriate when it pressures a harmed person to negotiate with an alleged perpetrator without safeguards.

8. Applied Case: An Inclusive Model Review Meeting

The following is an illustrative case.

Background

A university AI team is developing a system to prioritise applications for a public service. The team includes a project manager, two ML engineers, a domain expert, a community representative and a student research assistant who uses captions because of hearing loss.

Problem

During a review meeting, the project manager speaks rapidly and refers repeatedly to “common sense.” The community representative raises concerns that the training data underrepresent migrants and rural applicants. An engineer interrupts and says the concerns are “political.” The student research assistant misses several comments because captions are delayed. The manager concludes that the model is ready for pilot deployment.

Analysis

The problem is interpersonal, technical and ethical. The meeting has accessibility barriers, unequal participation and an attempt to dismiss relevant evidence through status language. The phrase “common sense” hides assumptions that may not be shared across communities. The decision also suffers from inadequate challenge and premature closure.

Approach and solution

The project lead pauses the decision and states: “We have not yet heard all evidence, and the meeting format is not allowing equal participation.” The team then:

• provides the agenda and model documentation in advance;
• enables reliable live captions and a written chat channel;
• gives each participant uninterrupted time to identify risks;
• asks the community representative to specify affected populations and evidence;
• records subgroup performance and data limitations;
• distinguishes technical disagreement from personal judgement;
• schedules a separate review of fairness, privacy and governance;
• defines a deployment gate requiring documented evidence and stakeholder sign-off.

The engineer revises the response to: “I interpret the concern differently. The current validation results show acceptable average performance, but I agree that subgroup results are needed before we generalise.”

Outcome

The pilot is delayed. Subsequent analysis reveals materially lower performance for one underrepresented group. The team expands data collection, consults affected communities and changes its communication protocol. The delay increases short-term workload but prevents an unjustified deployment and improves trust.

Learning points

1. Inclusion improves technical reasoning by broadening the evidence considered.

2. Accessibility is a condition of meaningful participation, not a courtesy.

3. Average performance can conceal subgroup harms.

4. Assertive disagreement is compatible with respect.

5. Leaders are responsible for meeting structures, not merely for inviting attendance.

9. Digital Communication and AI-Mediated Interaction

Email, messaging platforms, collaborative documents and video meetings alter interpersonal communication. Written messages preserve context but may remove tone and immediate repair. A short message can appear efficient to its author and hostile to its recipient. Before sending a sensitive message, check audience, purpose, tone, privacy and permanence.

Use descriptive subject lines, accessible formatting and explicit requests. Do not use group channels to shame individuals. Record decisions, but distinguish confirmed decisions from tentative ideas. In meetings, do not assume that silence means agreement.

Generative AI tools can assist with drafting, translation, summarisation and accessibility. However, AI-generated text may reproduce stereotypes, mistranslate identity-related language, expose confidential information or create a false impression of empathy. Users remain responsible for reviewing output, protecting personal data, obtaining consent where required and preserving human accountability. AI should support—not replace—listening, judgement and relationship repair.

10. Professional, Ethical and Organisational Implications

Inclusive communication is connected to professional duties. It supports accurate requirements, safer products, fairer evaluation and responsible governance. Organisations should establish clear conduct policies, accessible reporting channels, anti-retaliation protections, training and transparent investigation procedures. Policies are ineffective if leaders ignore violations or place the burden entirely on those harmed.

AI professionals should also consider the communication of uncertainty. Presenting a model as objective, infallible or universally accurate can mislead stakeholders. Ethical communication reports limitations, affected populations, confidence boundaries and unresolved risks in language appropriate to the audience. Transparency is not achieved by overwhelming non-specialists with technical detail; it requires relevant, understandable and honest explanation.

Challenges and Limitations

Communication frameworks are aids, not mechanical solutions. Empathy can be difficult when accounts conflict. Cultural generalisations may accidentally reproduce stereotypes. Assertiveness may be perceived differently across status or cultural contexts. “Neutral” language can conceal structural inequality, while excessive concern about politeness can silence legitimate criticism.

Individual skill cannot solve organisational problems such as discriminatory incentives, insecure employment or inaccessible systems. Responsibility must therefore be shared among individuals, educators, managers, institutions and technology designers. Inclusion should be evaluated through participation, outcomes and lived experience rather than through statements of intention alone.

Best-Practice Checklist

Before communicating, ask:

• What is my purpose, and who may be affected?
• What assumptions am I making?
• Is the language specific, respectful and accessible?
• Does the channel fit the sensitivity and urgency of the issue?
• Have I allowed meaningful opportunity for response?
• Am I distinguishing evidence from interpretation?
• Have I considered power, identity and accessibility?
• What action will follow if harm or disagreement is identified?

Chapter Summary

Interpersonal communication creates both informational and relational meaning. Emotional intelligence supports communication through self-awareness, self-management, empathy and relationship management. Assertiveness enables people to express needs and disagreement clearly without aggression or passivity. Active listening, constructive feedback, reliable follow-through and repair of harm strengthen trust.

Inclusive communication recognises that ability, identity, culture, language and power shape participation and interpretation. Accessibility, respectful identity language, plain language, equitable meeting practices and attention to power are essential in diverse AI and ML teams. Stereotypes, microaggressions and discriminatory language can damage individuals, decision quality and institutional legitimacy. Conflict should be managed through evidence, listening and shared interests, while serious misconduct requires formal safeguards. Digital and AI-mediated communication offer useful tools but do not remove human responsibility. Inclusive communication is therefore both a professional capability and an ethical requirement for responsible AI practice.

Key Terms

Active listening: Deliberate attention, interpretation, checking and response to another person’s message.

Assertiveness: Direct and respectful expression of needs, views and boundaries.

Cognitive empathy: Understanding another person’s perspective or likely experience.

Communication context: The social, cultural, organisational and technological conditions surrounding interaction.

Discrimination: Unequal treatment, opportunity or outcome associated with a social identity or group membership.

Emotional intelligence: Capacities for understanding and managing emotions in oneself and interaction.

Inclusive communication: Communication designed to enable respectful and meaningful participation across differences.

Intersectionality: The way multiple identities and systems of power combine to shape experience.

Microaggression: A commonplace message or action that communicates exclusionary or derogatory meaning, often indirectly.

Non-verbal communication: Meaning conveyed through behaviour, expression, posture, voice and interactional cues.

Passivity: Avoidance of expressing legitimate needs, views or boundaries.

Power: The capacity to influence decisions, resources, participation or consequences.

Positionality: The social and organisational position from which a person communicates and is interpreted.

Prejudice: A favourable or unfavourable attitude toward a person or group based on group membership.

Relationship management: Building trust, coordinating expectations and handling interaction constructively.

Self-awareness: Recognition of one’s emotions, assumptions, habits, position and effects on others.

Stereotype: A simplified generalisation about a social group.

Transactional communication: Communication in which participants simultaneously send, receive and interpret messages.

Review Questions

Short Answer Questions

1. What is the difference between empathy and agreement?

2. Define assertive communication and distinguish it from aggressive communication.

3. Identify two practices that support accessibility in meetings.

4. What is a microaggression?

5. Why can power affect participation in a meeting?

Long Answer Questions

1. Explain the dimensions of emotional intelligence and show how they support teamwork in an AI project.

2. Compare passive, aggressive, passive-aggressive and assertive communication, using examples.

3. Discuss how culture and language can influence communication without assuming that all members of a culture communicate identically.

4. Explain the role of active listening and constructive feedback in relationship management.

5. Discuss the ethical responsibilities of AI professionals when communicating uncertainty and model limitations.

Analytical / Application Questions

1. A senior engineer repeatedly interrupts a junior colleague during a model review. Analyse the power and communication issues and propose an appropriate intervention.

2. Design an inclusive meeting protocol for a multilingual AI team that includes a participant requiring captions and remote participation.

3. A team claims that its model is fair because its average accuracy is high. Using the principles in this chapter, formulate questions that would test the claim and improve the team’s communication and decision-making.

Multiple Choice Questions with Answers

Question 1

Which statement best describes assertive communication?

A. Avoiding disagreement to preserve harmony

B. Expressing needs clearly while respecting others

C. Using pressure to obtain compliance

D. Communicating resistance indirectly

Correct Answer: B. Expressing needs clearly while respecting others

Question 2

What is the primary purpose of active listening?

A. To prepare a rebuttal quickly

B. To demonstrate agreement with every statement

C. To check understanding and respond thoughtfully

D. To reduce the length of a conversation

Correct Answer: C. To check understanding and respond thoughtfully

Question 3

Which example is most clearly a stereotype?

A. Asking a colleague how they prefer to receive feedback

B. Documenting a decision after a meeting

C. Assuming that a group shares one fixed communication style

D. Providing captions for a recorded lecture

Correct Answer: C. Assuming that a group shares one fixed communication style

Question 4

Which practice most directly supports inclusive meetings?

A. Allowing the most senior person to speak first and longest

B. Using unexplained technical abbreviations

C. Sharing materials in advance and offering accessible participation channels

D. Treating silence as automatic agreement

Correct Answer: C. Sharing materials in advance and offering accessible participation channels

Question 5

What does self-awareness involve?

A. Suppressing all emotional responses

B. Recognising one’s emotions, assumptions and effects on others

C. Avoiding feedback from colleagues

D. Assuming that one’s intentions determine impact

Correct Answer: B. Recognising one’s emotions, assumptions and effects on others

Question 6

Which feedback statement is most constructive?

A. You are careless with data.

B. Everyone knows you cannot present well.

C. The evaluation population was not reported, so generalisability is unclear; please add it in the revision.

D. Your work is bad, but perhaps you will improve.

Correct Answer: C. The evaluation population was not reported, so generalisability is unclear; please add it in the revision.

Question 7

Why is the phrase “common sense” potentially problematic in a diverse team?

A. It always refers to a mathematical principle

B. It may conceal assumptions that are not shared across communities

C. It prevents all technical discussion

D. It is inaccessible only in written communication

Correct Answer: B. It may conceal assumptions that are not shared across communities

Question 8

Which response is most appropriate when a serious harassment allegation involves a substantial power imbalance?

A. Require the affected person to resolve it privately with the alleged perpetrator

B. Ignore it unless the team’s productivity declines

C. Use appropriate formal procedures and safeguards

D. Discuss the allegation publicly before investigating

Correct Answer: C. Use appropriate formal procedures and safeguards

Question 9

What is a responsible use of generative AI in interpersonal communication?

A. Sending generated messages without reviewing them

B. Using it to replace human listening and judgement

C. Using it for drafting while checking accuracy, privacy and tone

D. Entering confidential complaints to obtain a more empathetic reply

Correct Answer: C. Using it for drafting while checking accuracy, privacy and tone

Question 10

Why should AI teams examine subgroup performance rather than only average performance?

A. Average performance is never useful

B. Subgroup analysis can reveal unequal harms hidden by an aggregate result

C. Subgroup performance always produces lower accuracy

D. Technical models cannot be evaluated using averages

Correct Answer: B. Subgroup analysis can reveal unequal harms hidden by an aggregate result