ALEX Institute / AI / AI Ethics · Module 4

Stakeholder engagement and impact assessment

Track C · AI Ethics, module 4. Turing Commons · about 31 min reading (6,109 words).

OPEN Source and license. “AI Ethics and Governance” by The Alan Turing Institute, Turing Commons (Christopher Burr and contributors). Source: Turing Commons, AI Ethics and Governance skills track. License: CC BY 4.0. Turing Commons states its material is available under CC BY 4.0 (README and the track's welcome page). Changes: images and interactive elements removed; headings and links adapted for this page.

← Track C · AI Ethics: all modules

AI Sustainability and Stakeholder Engagement

Chapter Outline

Chapter Summary

In this chapter we look at the importance of stakeholders in sustainability through what is known as the stakeholder engagement processes (SEP) as well as the stakeholder impact assessment (SIA). The former is a three-staged process with the following core steps: (i) stakeholder analysis and salience, (ii) positionality reflection, and (iii) stakeholder engagement objective and method selection.

The latter goes beyond the SEP and is designed to ensure a project can not only remain sustainable, but also support the sustainability of impacted communities throughout time. Beyond development and initial deployment, an AI system requires continuous monitoring to ensure that the long-term transformative effects and impacts of the project are not only well-understood, but accepted and beneficial for impacted individuals and communities.

Learning Objectives

In this chapter, you will:

  • Familiarise yourself with the concept of sustainability in the context of AI and the importance of anticipatory reflection.
  • Learn about the AI project lifecycle and how it can be a useful tool for AI sustainability.
  • Understand the importance of an appropriate engagement with stakeholders, as well as how to conduct a stakeholder engagement process (SEP).
  • Familiarise yourself with impact assessments, in particular a stakeholder impact assessment (SIA) as well as its main components.

Stakeholder Engagement Process

A crucial component of AI Sustainability is conducting an appropriate Stakeholder Engagement Process (SEP). But before we delve into the process itself, we must ask, what is a stakeholder?

Key Concept: Stakeholder

Stakeholders are individuals or groups that:

(1) have interests or rights that may be affected by the past, present, and future decisions and activities of an organisation;

(2) may have the power or authority to influence the outcome of such decisions and activities;

(3) possess relevant characteristics that put them in positions of advantage or vulnerability with regard to those decisions and activities.

The SEP is an iterative process with three core steps:

Each of these activities should be documented and utilised to create a Project Summary (PS) report. A Project Summary Report is comprised of four components reflecting the SEP process: (1) a preliminary project scoping and stakeholder analysis, (2) a positionality reflection (3) an overview of established stakeholder objectives and methods and (4) a governance workflow map. The fourth component of the PS Report will be discussed later in the course when we look into accountability and governance of AI systems.

We will now analyse each of the relevant steps within the process and look at how they relate to thinking about the sustainability of the project as a whole.

Preliminary Project Scoping and Stakeholder Analysis

This is the first activity within the SEP process. It involves four sub-steps:

  1. Outlining project use context, domain, and data: In this step, a high-level description of the prospective system is outlined. This includes the domain in which it will operate, the contexts in which it will be used, and the data on which it will be trained. During this initial project scoping activity, information should be drawn from organisational documents (i.e., the project business case, proof of concept, or project charter), project team collaboration, as well as desk research (if necessary) to complete the description.

  2. Identifying stakeholders: Once the first step is completed and building on this contextual understanding, this step involves identifying who may be affected by, or may affect the project.

  3. Scoping potential stakeholder impacts: This involves carrying out a preliminary evaluation of the potential impacts of the prospective AI system on affected individuals and communities. At this initial stage of reflection, members of the project team should review the SUM values, and the corresponding ethical concerns, rights and freedoms, and then consider which of these could be impacted by the design, development and deployment of the prospective AI system and how.

  4. Analysing stakeholder salience: The step requires assessing the relevance of each identified stakeholder group to the AI project and to its use contexts. This includes assessing the relative interests, rights, vulnerabilities, and advantages of identified stakeholders as these interests, rights, vulnerabilities, and advantages may be impacted by, or may impact, the AI system in question. When identifying stakeholders, the project team should also consider organisational stakeholders, whose input will likewise strengthen the openness, inclusivity, and diversity of the project.

Stakeholder analyses may be carried out in a variety of ways that involve more-or-less stakeholder involvement. This spectrum of options ranges from analyses carried out exclusively by a project team without active community engagement to analyses built around the inclusion of community-led participation and co-design from the earliest stages of stakeholder identification. The degree of stakeholder involvement will vary from project to project based upon a preliminary assessment of the potential risks and hazards of the model or tool under consideration. 

Low-stakes AI applications that are not safety-critical, do not directly impact the lives of people, and do not process potentially sensitive social and demographic data may need less proactive stakeholder engagement than high-stakes projects. For example, an application which sorts out email and identifies spam may be correctly classified as low impact and low risk.

A responsible and thorough initial evaluation of the scope of the possible risks that could arise from the AI system (to be carried by the team responsible for developing said system) will determine the the potential hazards the project poses to affected stakeholders (be them individuals or communities). A reasonable assessments of the dangers posed to individual wellbeing and public welfare is needed in order to formulate proportionate approaches to stakeholder involvement. 

Regardless of the potential impacts of a project, involving affected individuals and communities in stakeholder analysis (and, later, in stakeholder impact assessment) should, in all cases, be a significant consideration. Stakeholder involvement ensures that a project will possess an appropriate degree of public accountability, transparency, legitimacy, and democratic governance, and it recognizes the important role played in this by the inclusion of the voices of all affected individuals and communities in decision-making processes. 

In addition to providing these important supports for building public trust, stakeholder involvement can help to strengthen the objectivity, reflexivity, reasonableness, and robustness of the choices a project team makes across the project lifecycle. This is because the inclusion of a wider range of perspectives (especially of those who are most marginalised) can enlarge a project team’s purview, expand its domain knowledge as well as its understanding of citizens’ needs. It can likewise unearth potential biases that may arises from limiting the standpoints that inform decision-making to those of team members. 

Public engagement and community involvement, however, are only one part of the measures a team needs to take to ensure the objectivity, reflexivity, reasonableness, and robustness of its stakeholder analysis, impact assessment, and decision-making, more generally. Apart from outward-facing community participation, processes inward-facing reflection should also inform the way the team approaches to these challenges (see the sustainability, engagement, and reflection cycle).

Positionality Reflection

All individual human beings come from unique places, experiences, and life contexts that have shaped their thinking and perspectives. Reflecting on these is important insofar as it can help team members understand how their viewpoints might differ from those around them and, more importantly, from those who have diverging cultural and socioeconomic backgrounds and life experiences.

Identifying and probing these differences can enable individuals to better understand how their own backgrounds, for better or worse, frame the way they see others, the way they approach and solve problems, and the way they carry out research and engage in innovation. By undertaking such efforts to recognise social position and differential privilege, they may gain a greater awareness of their own personal biases and unconscious assumptions. This then can enable them to better discern the origins of these biases and assumption and to confront and challenge them in turn. 

Social scientists have long referred to this kind of self-locating reflection as “positionality”. When team members take their own positionalities into account making them explicit, they can better grasp how the influence of their respective social and cultural positions creates strengths and limitations.

On the one hand, one’s positionality—with respect to characteristics like ethnicity, race, age, gender, socioeconomic status, education and training levels, values, geographical background, etc.—can have a positive effect on an individual’s contributions to an innovation project; the uniqueness of each person’s lived experience and standpoint can play a constructive role in introducing insights and understandings that other team members do not have. On the other hand, one’s positionality can assume a harmful role when hidden biases and prejudices that derive from a person’s background, and from power imbalances and differential privileges, creep into decision-making processes undetected. 

When taking positionality into account, team members should reflect on their own positionality matrix. They should ask: To what extent do my personal characteristics, group identifications, socioeconomic status, educational, training, & work background, team composition, & institutional frame represent sources of power and advantage or sources of marginalisation and disadvantage? How does this positionality influence my (and my team’s) ability to identify & understand affected stakeholders and the potential impacts of my project? Several other questions must be asked to respond to these two: How do I identify? How have I been educated and trained? What does my institutional context and team composition look like? What is my socioeconomic history? 

Stakeholder Engagement Objective and Method

Once the initial project scoping, stakeholder analysis, and positionality reflection have been done, a project team can move on to define the objective for their stakeholder engagement, as well as the method(s) to be used.

As we have previously established, stakeholder engagement may be carried out in a variety of ways that involve more-or-less stakeholder involvement. This spectrum of options ranges from analyses carried out exclusively by a project team without active community engagement to analyses built around the inclusion of community-led participation and co-design from the earliest stages of stakeholder identification.

The degree of stakeholder involvement will vary from project to project and take into account a preliminary assessment of the potential risks and hazards of the model or tool under consideration. Low-stakes AI applications that are not safety-critical, do not directly impact the lives of people, and do not process potentially sensitive social and demographic data may need less proactive stakeholder engagement than high-stakes projects. Similarly, the project team will need to take into account their assessment of positionality.

When weighing these three factors, the team should prioritise the establishment of a clear and explicit stakeholder engagement objective and document this. This is crucial, because all stakeholder engagement processes can run the risk either of being cosmetic tools employed to legitimate projects without substantial and meaningful participation or of being insufficiently participative, i.e. of being one-way information flows or nudging exercises that serve as public relations instruments. The purpose of stakeholder involvement in sustainable AI projects is just the opposite: to amplify the participatory agency of affected individuals and organisations in impact assessment, risk management, and assurance processes.

To avoid such hazards of superficiality, the team should shore up its proportionate approach to stakeholder engagement with deliberate and precise goal-setting. There are a range of engagement options that can help a project obtain a level of public participation which meets team-based assessments of impact and positionality as well as practical considerations and stakeholder needs.

Defining a Stakeholder Objective

Selecting a Stakeholder Engagement Method

The following table summarises a wide range of salient methods:

Mode of Engagement Description Degree of Engagement Practical Strengths Practical Weaknesses
:material-email: Newsletters (email) Regular emails (e.g.: fortnightly or monthly) that contain updates, relevant news, and calls to action in an inviting format. INFORM Can reach many people; can contain large amount of relevant information; can be made accessible and visually engaging. Might not reach certain portions of the population; can be demanding to design and produce with some periodicity; easily forwarded to spam/junk folders without project team knowing (leading to overinflated readership stats).
:material-mailbox: Letters (post) Regular letters (e.g.: monthly) that contain the latest updates, relevant news and calls to action. INFORM Can reach parts of the population with no internet or digital access; can contain large amount of relevant information; can be made accessible and visually engaging. Might not engage certain portions of the population; Slow delivery and interaction times hampers the effective flow of information and the organisation of further engagement.
:material-checkbox-blank-badge: App notifications Projects can rely on the design of apps that are pitched to stakeholders who are notified on their phone with relevant updates. INFORM Easy and cost-effective to distribute information to large numbers of people; Rapid information flows bolster the provision of relevant and timely news and updates. More significant initial investment in developing an app; will not be available to people without smartphones.
:fontawesome-solid-people-group: Community fora Events in which panels of experts share their knowledge on issues and then stakeholders can ask questions. INFORM Can inform people with more relevant information by providing them with the opportunity to ask questions; brings community together in a shared space of public communication. More time-consuming and resource intensive to organise; might attract smaller numbers of people and self-selecting groups rather than representative subsets of the population; effectiveness is constrained by forum capacity.
:fontawesome-solid-list-check: Online Surveys Survey sent via email, embedded in a website, shared via social media, etc. CONSULT Cost-effective; simple mass- distribution. Risk of pre-emptive evaluative framework when designing questions; Does not reach those without internet connection or computer/smartphone access.
:material-phone-in-talk: Phone Interviews Structured or semi-structured interviews held over the phone. CONSULT PARTNER Opportunity for stakeholders to voice concerns more openly. Risk of pre-emptive evaluative framework when designing questions; Might exclude portions of the populations without phone access or with habits of infrequent phone use.
:fontawesome-solid-door-closed: Door-to-door interviews Structured or semi-structured interviews held in-person at people’s houses. CONSULT PARTNER Opportunity for stakeholders to voice concerns more openly; can allow participants the opportunity to form connections through empathy and face-to- face communication. Potential for limited interest to engage with interviewers; time-consuming; can be seen by interviewees as intrusive or burdensome.
:fontawesome-solid-people-arrows-left-right: In-person interviews Short interviews conducted in- person in public spaces. CONSULT PARTNER Can reach many people and a representative subset of the population if stakeholders are appropriately defined and sortition is used. Less targeted; pertinent stakeholders must be identified by area; little time/interest to engage with interviewer; can be viewed by interviewees as time- consuming and burdensome.
:material-selection-search: Focus Groups A group of stakeholders brought together and asked their opinions on a particular issue. Can be more or less formally structured. CONSULT PARTNER Can gather in-depth information; Can lead to new insights and directions that were not anticipated by the project team. Subject to hazards of group think or peer pressure; complex to facilitate; can be steered by dynamics of differential power among participants.
:fontawesome-solid-users-between-lines: Online Workshops Workshops using digital tools such as collaborative platforms. CONSULT Opportunity to reach stakeholders across regions, increased accessibility depending on digital access. Potential barriers to accessing tools required for participation, potential for disengagement.
:material-crowd: Crowdsourcing (Online) Well-designed tasks that can be undertaken by a distributed collective, with individuals working on separate components. CONSULT PARTNER Opportunity to engage stakeholders across regions, in an asynchronous manner, with increased accessibility depending on digital access. Supports increased potential for diverse forms of expertise and experience. Can be misused as a method for outsourcing cheap labour; potential barriers to accessing tools required for participation; potential for disengagement; difficult to ensure accuracy and validity of input from participants.
:material-github: Distributed Project Collaboration (Online) Online digital platforms, such as GitHub, enable new forms of citizen science and collaborative development on diverse projects (e.g., open source software, open science). CONSULT PARTNER EMPOWER Opportunity to engage stakeholders across regions, in an asynchronous manner, with increased accessibility depending on digital access; increased potential for diverse forms of expertise and experience; empowers new communities to actively participate in shaping and building tools that have real value for their communities. Potential barriers to accessing digital tools required for participation, including high levels of digital literacy.
:fontawesome-solid-building-columns: Citizen panel or assembly Large groups of people (dozens or even thousands) who are representative of a town/region. INFORM CONSULT PARTNER EMPOWER Provides an opportunity for co-production of outputs; can produce insights and directions that were not anticipated by the project team; can provide an information base for conducting further outreach (surveys, interviews, focus groups, etc.); can be broadly representative; can bolster a community’s sense of democratic agency and solidarity. Participant rolls must be continuously updated to ensure panels or assemblies remains representative of the population throughout their lifespan; resource-intensive for establishment and maintenance; subject to hazards of group think or peer pressure; complex to facilitate; can be steered by dynamics of differential power among participants.
:material-account-group: Citizen jury A small group of people (between 12 and 24), representative of the demographics of a given area, come together to deliberate on an issue (generally one clearly framed set of questions), over the period of 2 to 7 days (Involve.org.uk. INFORM CONSULT PARTNER EMPOWER Can gather in-depth information; can produce insights and directions that were not anticipated by the project team; can bolster participants’ sense of democratic agency and solidarity. Subject to hazards of group think; complex to facilitate; risk of pre-emptive evaluative framework; small sample of citizens involved risks low representativeness of wider range of public opinions and beliefs.

As with all forms of engagement, deciding on the best method requires awareness of your audience. Consider the following cases:

":octicons-log-16: Policymakers"

A research team has released results from an economics study that could have a positive impact on public policy. They decide to share these results with policymakers. The goal is to directly influence policy. Therefore, the results need to be clearly communicated and also connected to the policy goal. This connection is important to help ensure that policy-makers are able to evaluate the wider implications of the scientific findings.

**Communication Goal:** to demonstrate how scientific findings can support evidence-based policy impact

":octicons-people-16: General Public"

As part of an education outreach campaign to improve digital literacy among adolescents, a mental health charity are running workshops with secondary school students. They wish to communicate recent evidence about the impact of over-using social media on mental health. Rather than communicating complicated statistical information about the methods used in their study, the team develop a more accessible form of their findings and link these findings to practical steps that the students can take to protect themselves online.

**Communication Goal:** to build awareness of possible risks associated with excessive social media usage and support behavioural change strategies

":octicons-beaker-16: Researchers"

A PhD student working in a Physics department has results from a recent study that developed and tested a new method for the large-scale data mining of astronomical data. The PhD student wishes to present this new method and the validation study at an upcoming international conference for data science. The audience will be technically literate, but will not have specialist expertise in astronomy. Therefore, the PhD student describes the method in the context of its original study buy also emphasises the generalisability for other sciences (e.g. genomics).

**Communication Goal**: to advance academic career by gaining experience of presenting conference papers and also generating interest in a novel data science method.
Information is taken from PSGW on Sustainability, Part 1, plus Public Engagement course on Turing Commons

Stakeholder Impact Assessment

Designers and users of AI systems should remain aware that these technologies may have transformative and long-term effects on individuals and society. To ensure that the deployment of an AI system remains sustainable and supports the sustainability of the communities it will affect, the project team should proceed with a continuous sensitivity to the real-world impacts that the system will have. The team should come together to evaluate the social impact and sustainability of an AI project through what is known as a Stakeholder Impact Assessment (SIA).

The SUM values form the basis of the SIA. They are intended as a launching point for open and inclusive conversations about the individual and societal impacts of data science research and AI innovation projects rather than to provide a comprehensive inventory of moral concerns and solutions. At the very outset of any project, these should provide the normative point of departure for collaborative and anticipatory reflection, while, at the same time, allowing for the respectful and interculturally sensitive inclusion of other points of view.

Objectives of a SIA

The purpose of carrying out a SIA is multidimensional. SIAs can serve several purposes, some of which include:

- Help to build **public confidence** that the design and deployment of the AI system has been done responsibly.
- Facilitate and strengthen your **accountability** framework.
- **Bring to light unseen risks** that threaten to affect individuals and the public good.
- Underwrite **well-informed decision-making** and **transparent innovation practices**.
- Demonstrate **forethought** and due **diligence** not only within an organisation but also to the wider public.

AI projects may require different kinds of impact assessments. For example Data Protection Law requires data protection impact assessments (DPIAs) to be carried out in cases where the processing of personal data is likely to result in a high risk to individuals. DPIAs assess the necessity and proportionality of the processing of personal data, identify risks that may emerge in that processing, and present measures taken to mitigate those risks. Another example are equality impact assessments (EIAs) which aid in fulfiling the requirements of equality legislation.

While both DPIAs and EIAs provide relevant insights with respect to the ethical stakes of AI innovation projects, they go only part of the way in identifying and assessing the full range of potential individual and societal impacts of the design, development, and deployment of AI systems. Reaching a comprehensive assessment of these impacts is the purpose of SIAs.

Key Concept: Stakeholder Impact Assessment

SIAs are tools that create a procedure for, and a means of documenting, the collaborative evaluation and reflective anticipation of the possible harms and benefits of AI innovation projects. SIAs are not intended to replace DPIAs or EIAs, which are obligatory. Rather, SIAs are meant to be integrated into the wider impact assessment regime as a way to demonstrate that sufficient attention has been paid to the ethical permissibility, transparency, accountability, and equity of AI innovation projects.

There are three critical points in the AI project lifecycle at which the project team should convene to impact the social impact and sustainability of a project:

":octicons-log-16: Design"

A SIA should be carried out to determine the ethical permissibility of the project. As a starting point, the team should refer to the SUM Values  for the considerations of the possible effects of the project on individual wellbeing and public welfare. This should include a stakeholder engagement and involvement component in the initial SIA through methods established in the initial Project Summary Report (PS Report), so public views can be considered in ways that are proportional to potential project impacts and appropriate to team positionality.

This will bolster the inclusion of a diversity of voices and opinions into the design and development process through the participation of a more representative range of stakeholders. The Design Phase SIA includes a revisitation of the Project Summary Report, where engagement objectives and methods for the Development Phase SIA were first established. These, and other relevant project revisions should be reflected in an update to the PS Report.

":material-developer-board: Develop"

Once a model has been trained, tested, and validated, the project team should revisit the initial SIA to confirm that the AI system to be implemented is still in line with the evaluations and conclusions of the original assessment. This check-in should be logged on the Development Phase section of the SIA with any applicable changes added and discussed.

The method of stakeholder engagement that accompanies the SIA process will have been initially established in the PS report and revisited in the Design Phase SIA. This report should be revisited again during the Development Phase SIA and updated where needed. At this point the team must also set a time frame for re-assessment once the system is in operation. Time frames for these re-assessments should be decided by the team on a case-by-case basis but should be proportional to the scale of the potential impact of the system on the individuals and communities it will affect.

":material-network-pos: Deploy"

Once an AI system has gone live, the team should iteratively revisit and re-evaluate the SIA. These check-ins should be logged on the Deployment Phase section of the SIA with any applicable changes added and discussed. Deployment-Phase SIAs should focus both on evaluating the existing SIA against real world impacts and on considering how to mitigate the unintended consequences that may have ensued in the wake of the deployment of the system. As with each SIA iteration, the PS report is revisited at this point, when objectives, methods, and time frames for the next Deployment Phase SIA are established.

Skills for Conducting SIA's

Weighing the values and considering trade-offs

Taking the SUM values as a starting point of conversation and for the SIA, there will come times when these values come into conflict with one another and decisions will have to be made on which value to prioritise. Within a team, discussion should be encouraged on how to weigh the values against one another and how to consider trade-offs should use-case specific circumstances arise when the values come into tension with each other.

For instance, there may be circumstances where the use of an AI system could optimally advance the public interest only at the cost of safeguarding the wellbeing or the autonomy of a given individual. In other cases, the use of an AI system could preserve the wellbeing of a particular individual only at the cost of the autonomy of another or of the public welfare more generally.

This issue of adjudicating between conflicting values has long been a crucial and thorny dimension of collective life, and the problem of discovering reasonable ways to overcome the disagreements that arise as a result of the plurality of human values has occupied thinkers for just as long. Nonetheless, over the course of the development of modern democratic and plural societies, several useful approaches to managing the tension between conflicting values have emerged.

Consequences-based and principles-based approaches to balancing values

Let's go back to some concepts introduced in chapter 1 that will be useful when trying to balance the tension between values: consequences-based moral thinking or consequentialism and principles-based moral thinking or deontology. These can be seen as procedural tools for thinking through a given dilemma in weighing values.

As a quick reminder, a consequence-based approach asks that, in judging the moral correctness of an action, one prioritise considerations of the goodness produced by an outcome. In other words, because what matters most is the consequences of ones actions, the goodness of these consequences should be maximised. Standards of right and wrong (indicators of what one ought to do) are determined, on this view, by whether the action taken maximises overall goodness of the consequences rather than by the principles or standards one applies when acting.

A principles-based approach takes the opposite tack. The rightness of an action is determined, from this standpoint, by the intentional application of a universally applicable standard, maxim, or principle. Rather than basing the morality of conduct on the ends served by it, this approach anchors rightness in the duty or obligation of the individual agent to follow a rationally determined (and therefore “universalisable”) principle.

For deontological or principles-based ethics, the integrity of the principled action and intention matters most, and so justified constraints, which are rooted in the priority to act according to moral standards, must be put on the pursuit of the achievement of one’s goals.

Knowing when to prioritise consequences and when to prioritise principles in moral deliberations is a tricky matter, and applying either a consequentialist or deontological approach (or both) may make sense depending upon the context.

To take a familiar example, in deontologically following the principle ‘Thou shall not lie,’ you would be justifiably constrained from stealthily deceiving and misleading others to get ahead in your job. The principle matters here. But, in a different situation, say, where lying to a murderer, who appears at your front door, would save an innocent victim whom you are concealing in your cellar, the prioritisation of consequences makes more sense.

Consider another example: In an overburdened council, the introduction of an automated system for making sites available for development would vastly expedite housing delivery. The implementation of this AI system would thus produce a consequence that could be beneficial to the public.

Yet, it may, among other things, simultaneously do damage to the value of Connect (which safeguards interpersonal dialogue, meaningful human connection and social cohesion) by eliminating time intensive consultation processes that facilitate interpersonal communication, trust building, and social bonding between council staff and residents. How then could one go about weighing the value of improving public welfare against the value of respecting the integrity of interpersonal relations?

One way would be to place each side of this comparison under the rubric of either consequences or principles and then measure them up against each other accordingly. From one perspective, the publicly beneficial consequences of improving service delivery might outweigh the publicly harmful consequences of impairing social cohesion. On a different view, such a trade-off would be unacceptable, because the principle of respecting the integrity of social cohesion trumps any solidarity-harming but publicly beneficial consequences whatsoever. The answers to these questions will always be tricky, but deliberation between team members and stakeholders should always be a part of arriving to any sort of consensus (which we will focus on in the next section).

Getting clear on the consequences and the principles involved in a specific case of conflicting values will allow team members to get a better picture of the practical and moral stakes at play in a particular project. It will also help the team get a sharper idea of the proportionality of using an AI technology to achieve a desired outcome given both its potential ethical impacts and the social needs to which it is responding. When drawn upon for guidance, consequentialism and deontology can provide a kind of procedural scale upon which to place, measure, and weigh conflicting values. They are practical tools that can be used within meaningful deliberation.

Ensuring meaningful and inclusive deliberation

The most general approach is to encourage mutually respectful, sincere, and well-informed dialogue, so that reasons from all affected voices can be heard and considered. Deliberations that have been inclusive, open, and impartial tend to generate better and more inferentially sound conclusions, so approaching the adjudication of conflicting values in this manner will likely improve mutual understanding of the rationales and perspectives, which inform those values.

Whether or not this ends up being the case in each conversational context, the importance of cultivating a culture of innovation, which encourages reciprocally respectful, open, non-coercive, and mutually accountable communication must be stressed. The success of the modern sciences (which have been built on the dynamic foundations of inclusive, rational, and democratic communication) is perhaps evidence enough to support the validity of this emphasis.

Here, procedural ethics is crucial. The importance of meaningful dialogue in balancing values is rooted in central role played in it by the rational exchange and assessment of ideas and beliefs. The validity of the claims we make in conversations about values is bound by practices of giving and asking for reasons. A claim about values that is justified is one that convinces by the unforced strength of the better or more compelling argument. Rational justification and persuasive reason-giving are, in fact, central elements of legitimate and consensus-oriented moral decision making. And, along the same lines, claims made about moral values or properties are subject to critical evaluation vis-à-vis their inferential strengths and weaknesses.

Another way to understand the importance of meaningful dialogue in balancing values has to do with the way that such a procedural view of ethical thinking enables open communication about prioritising values without imposing substantive views about the values themselves. Instead, an emphasis on rational communication in deliberations looks to secure a justified and equitable process of exchanging and evaluating reasons. It starts with the question: what are the enabling conditions in the interpersonal communication of values and beliefs that allow interlocutors to come to defensible and rationally acceptable moral judgments and reason-based consensus?

To answer this question, moral thinkers over the past century have endeavoured to reconstruct the practical assumptions behind and presuppositions of rational communication (a summary of the most essential of such assumptions and presuppositions is provided below). Creating a reflective and practicable awareness of these assumptions and presuppositions among members of a team can play a crucial role in creating an innovation environment that is optimally conducive meaningful and inclusive deliberation.

Preconditions of meaningful deliberation:

Impartiality

Interlocutors engaging in meaningful deliberation must consider the interests of all those who are affected by their actions equally. Thinking impartially involves taking on the view of others to try to put oneself in their place.

Non-coercion

Meaningful deliberation must be free from any sort of implicit or explicit coercion, force, or restriction that would prevent the open and unconstrained exchange of reasons.

Sincerity

Meaningful deliberation must be free from any sort of deception or duplicity that would prevent the authentic exchange of reasons. Interlocutors must mean what they say.

Consistency and coherence

Arguments and positions offered in meaningful deliberation must be clear, free from contradictions, and hold together collectively in an understandable way.

Mutual respect and egalitarian reciprocity

All interlocutors must be treated with respect and given equal opportunity to contribute to the conversation. All voices are worthy of equal consideration in processes of exchanging reasons.

Inclusivity and publicity

Anyone whose interests are affected by an issue and who could make a contribution to better understanding it must not be excluded from participating in deliberation. All relevant voices must be heard and all relevant information considered.

Addressing and mitigating power dynamics that may obstruct meaningful and inclusive deliberation

The stewardship of meaningful and inclusive dialogue is critical to safeguarding the collective weighing up of values. However, there is an important potential barrier to meaningful deliberation that challenges its feasibility and must be addressed. As guiding assumptions of rational communication, norms like sincerity, impartiality, non-coercion, and inclusiveness may strike some as overly idealistic. In the real world, discussions are rarely fully inclusive, informed, and free of assertive manipulation, coercion, and deception. Rather, deliberation and dialogue are often steered by and crafted to protect the interests of the dominant. Likewise, differential power relationships (for instance, divergent educational backgrounds that derive from differential socioeconomic privileges) create power imbalances that fundamentally challenge the conditions of reciprocity and equal footing that are needed for justified and equitable communication.

A team should confront these obstacles to meaningful deliberation head-on through a power-aware approach to facilitating collaborative reflection, dialogue, and engagement. An awareness of and sensitivity to the differential relationships of power that can suppress the full participation of disadvantaged or marginalised voices can better encourage an inclusive, open, and equal opportunity conversation between participants.

Clear-headed explorations of power dynamics between civil servants, scientists, citizens, domain experts, and policymakers can assist in avoiding the kind of deficiencies of representation and empowerment that risk reinforcing existing power structures and inequalities. This may involve active mitigation measures like the provision of training, upskilling, and technical resources to those who have lacked access to them. Above all, the norms of meaningful deliberation makes teams aware of the possible distortions of communication (i.e. a lack of egalitarian reciprocity, non-coercion, sincerity, etc.) that must be tackled and rectified for the hurdles of power disparities to be scaled.

← Track C · AI Ethics: all modules


ALEX Institute