contestability

On mapping AI value chains

At CSCW 2024, back in November of last year, we* ran a workshop titled “From Stem to Stern: Contestability Along AI Value Chains.” With it, we wanted to address a gap in contestable AI research. Current work focuses mainly on contesting specific AI decisions or outputs (for example, appealing a decision made by an automated content moderation system). But we should also look at contestability across the entire AI value chain—from raw material extraction to deployment and impact (think, for example, of data center activists opposing the construction of new hyperscales). We aimed to explore how different stakeholders can contest AI systems at various points in this chain, considering issues like labor conditions, environmental impact, and data collection practices often overlooked in contestability discussions.

The workshop mixed presentations with hands-on activities. In the morning, researchers shared their work through short talks, both in person and online. The afternoon focused on mapping out where and how people can contest AI systems, from data collection to deployment, followed by detailed discussions of the practical challenges involved. We had both in-person and online participants, requiring careful coordination between facilitators. We wrapped up by synthesizing key insights and outlining future research directions.

I was responsible for being a remote facilitator most of the day. But Mireia and I also prepared and ran the first group activity, in which we mapped a typical AI value chain. I figured I might as well share the canvas we used for that here. It’s not rocket science, but it held up pretty well, so maybe some other people will get some use out of it. The canvas was designed to offer a fair bit of scaffolding for thinking through what decision points there are along the chain that are potentially value-laden.

AI value chain mapping canvas (licensed CC-BY 4.0 Mireia Yurrita & Kars Alfrink, 2024). Download PDF.

Here’s how the activity worked: We covered about 50 minutes doing a structured mapping exercise where participants identified potential contestation points along an AI value chain, using ChatGPT as an example case. The activity used a Miro board with a preliminary map showing different stages of AI development (infrastructure setup, data management, AI development, etc.). Participants first brainstormed individually for 10 minutes, adding value-laden decisions and noting stakeholders, harms, benefits, and values at stake. They then collaborated to reorganize and discuss the map for 15 minutes. The activity concluded with participants using dot voting (3 votes each) to identify the most impactful contestation sites, which were then clustered and named to feed into the next group activity.

The activity design drew from two main influences: typical value chain mapping methodologies (e.g., Mapping Actors along Value Chains, 2017), which usually emphasize tracking actors, flows, and contextual factors, and Wardley mapping (Wardley, 2022), which is characterized by the idea of a structured progression along an x-axis with an additional dimension on the y-axis.

The canvas design aimed to make AI system development more tangible by breaking it into clear phases (from infrastructure through governance) while considering visibility and materiality through the y-axis. We ultimately chose to use a familiar system (ChatGPT). This, combined with the activity’s structured approach, helped participants identify concrete opportunities for intervention and contestation along the AI value chain, which we could build on during the rest of the workshop.

I got a lot out of this workshop. Some of the key takeaways that emerged out of the activities and discussions include:

There’s a disconnect between legal and technical communities, from basic terminology differences to varying conceptions of key concepts like explainability, highlighting the need for translation work between disciplines.
We need to move beyond individual grievance models to consider collective contestation and upstream interventions in the AI supply chain.
We also need to shift from reactive contestation to proactive design approaches that build in contestability from the start.
By virtue of being hybrid, we were lucky enough to have participants from across the globe. This helped drive home to me the importance of including Global South perspectives and considering contestability beyond Western legal frameworks. We desperately need a more inclusive and globally-minded approach to AI governance.

Many thanks to all the workshop co-organizers for having me as part of the team and to Agathe and Yulu, in particular, for leading the effort.

* The full workshop team consisted of Agathe Balayn, Yulu Pi, David Gray Widder, Mireia Yurrita, Sohini Upadhyay, Naveena Karusala, Henrietta Lyons, Cagatay Turkay, Christelle Tessono, Blair Attard-Frost, Ujwal Gadiraju, and myself.

Participatory AI and ML engineering

In the first half of this year, I’ve presented several versions of a brief talk on participatory AI. I figured I would post an amalgam of these to the blog for future reference. (Previously, on the blog, I posted a brief lit review on the same topic; this talk builds on that.)

So, to start, the main point of this talk is that many participatory approaches to AI don’t engage deeply with the specifics of the technology. One such specific is the translation work engineers do to make a problem “learnable” by a machine (Kang, 2023). From this perspective, the main question to ask becomes, how does translation happen in our specific projects? Should citizens be involved in this translation work? If so, how to achieve this?

Before we dig into the state of participatory AI, let’s begin by clarifying why we might want to enable participation in the first place. A common motivation is a lack of democratic control over AI systems. (This is particularly concerning when AI systems are used for government policy execution. These are the systems I mostly look at in my own research.) And so the response is to bring the people into the development process, and to let them co-decide matters.

In these cases, participation can be understood as an enabler of democratic agency, i.e., a way for subjects to legitimate the use of AI systems (cf. Peter, 2020 in Rubel et al., 2021). Peter distinguishes two pathways: a normative one and a democratic one. Participation can be seen as an example of the democratic pathway to legitimation. A crucial detail Peter mentions here, which is often overlooked in participatory AI literature, is that normative constraints must limit the democratic pathway to avoid arbitrariness.

So, what is the state of participatory AI research and practice? I will look at each in turn next.

As mentioned, I previously posted on the state of participatory AI research, so I won’t repeat that in full here. (For the record, I reviewed Birhane et al. (2022), Bratteteig & Verne (2018), Delgado et al. (2023), Ehsan & Riedl (2020), Feffer et al. (2023), Gerdes (2022), Groves et al. (2023), Robertson et al. (2023), Sloane et al. (2020), and Zytko et al. (2022).) Elements that jump out include:

Superficial and unrepresentative involvement.
Piecemeal approaches that have minimal impact on decision-making.
Participants with a consultative role rather than that of active decision-makers.
A lack of bridge-builders between stakeholder perspectives.
Participation washing and exploitative community involvement.
Struggles with the dynamic nature of technology over time.
Discrepancies between the time scales for users to evaluate design ideas versus the pace at which systems are developed.
A demand for participation to enhance community knowledge and to actually empower them.

Taking a step back, if I were to evaluate the state of the scientific literature on participatory AI, it strikes me that many of these issues are not new to AI. They have been present in participatory design more broadly for some time already. Many of these issues are also not necessarily specific to AI. The ones I would call out include the issues related to AI system dynamism, time scales of participation versus development, and knowledge gaps between various actors in participatory processes (and, relatedly, the lack of bridge-builders).

So, what about practice? Let’s look at two reports that I feel are a good representation of the broader field: Framework for Meaningful Stakeholder Involvement by ECNL & SocietyInside, and Democratizing AI: Principles for Meaningful Public Participation by Data & Society.

Framework for Meaningful Stakeholder Involvement is aimed at businesses, organizations, and institutions that use AI. It focuses on human rights, ethical assessment, and compliance. It aims to be a tool for planning, delivering, and evaluating stakeholder engagement effectively, emphasizing three core elements: Shared Purpose, Trustworthy Process, and Visible Impact.

Democratizing AI frames public participation in AI development as a way to add legitimacy and accountability and to help prevent harmful impacts. It outlines risks associated with AI, including biased outcomes, opaque decision-making processes, and designers lacking real-world impact awareness. Causes for ineffective participation include unidirectional communication, socioeconomic barriers, superficial engagement, and ineffective third-party involvement. The report uses environmental law as a reference point and offers eight guidelines for meaningful public participation in AI.

Taking stock of these reports, we can say that the building blocks for the overall process are available to those seriously looking. The challenges facing participatory AI are, on the one hand, economic and political. On the other hand, they are related to the specifics of the technology at hand. For the remainder of this piece, let’s dig into the latter a bit more.

Let’s focus on translation work done by engineers during model development.

For this, I build on work by Kang (2023), which focuses on the qualitative analysis of how phenomena are translated into ML-compatible forms, paying specific attention to the ontological translations that occur in making a problem learnable. Translation in ML means transforming complex qualitative phenomena into quantifiable and computable forms. Multifaceted problems are converted into a “usable quantitative reference” or “ground truth.” This translation is not a mere representation of reality but a reformulation of a problem into mathematical terms, making it understandable and processable by ML algorithms. This transformation involves a significant amount of “ontological dissonance,” as it mediates and often simplifies the complexity of real-world phenomena into a taxonomy or set of classes for ML prediction. The process of translating is based on assumptions and standards that may alter the nature of the ML task and introduce new social and technical problems.

So what? I propose we can use the notion of translation as a frame for ML engineering. Understanding ML model engineering as translation is a potentially useful way to analyze what happens at each step of the process: What gets selected for translation, how the translation is performed, and what the resulting translation consists of.

So, if we seek to make participatory AI engage more with the technical particularities of ML, we could begin by identifying translations that have happened or might happen in our projects. We could then ask to what extent these acts of translation are value-laden. For those that are, we could think about how to communicate these translations to a lay audience. A particular challenge I expect we will be faced with is what the meaningful level of abstraction for citizen participation during AI development is. We should also ask what the appropriate ‘vehicle’ for citizen participation will be. And we should seek to move beyond small-scale, one-off, often unrepresentative forms of direct participation.

Bibliography

Birhane, A., Isaac, W., Prabhakaran, V., Diaz, M., Elish, M. C., Gabriel, I., & Mohamed, S. (2022). Power to the People? Opportunities and Challenges for Participatory AI. Equity and Access in Algorithms, Mechanisms, and Optimization, 1–8. https://doi.org/10/grnj99
Bratteteig, T., & Verne, G. (2018). Does AI make PD obsolete?: Exploring challenges from artificial intelligence to participatory design. Proceedings of the 15th Participatory Design Conference: Short Papers, Situated Actions, Workshops and Tutorial – Volume 2, 1–5. https://doi.org/10/ghsn84
Delgado, F., Yang, S., Madaio, M., & Yang, Q. (2023). The Participatory Turn in AI Design: Theoretical Foundations and the Current State of Practice. Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, 1–23. https://doi.org/10/gs8kvm
Ehsan, U., & Riedl, M. O. (2020). Human-Centered Explainable AI: Towards a Reflective Sociotechnical Approach. In C. Stephanidis, M. Kurosu, H. Degen, & L. Reinerman-Jones (Eds.), HCI International 2020—Late Breaking Papers: Multimodality and Intelligence (pp. 449–466). Springer International Publishing. https://doi.org/10/gskmgf
Feffer, M., Skirpan, M., Lipton, Z., & Heidari, H. (2023). From Preference Elicitation to Participatory ML: A Critical Survey & Guidelines for Future Research. Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, 38–48. https://doi.org/10/gs8kvx
Gerdes, A. (2022). A participatory data-centric approach to AI Ethics by Design. Applied Artificial Intelligence, 36(1), 2009222. https://doi.org/10/gs8kt4
Groves, L., Peppin, A., Strait, A., & Brennan, J. (2023). Going public: The role of public participation approaches in commercial AI labs. Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, 1162–1173. https://doi.org/10/gs8kvs
Kang, E. B. (2023). Ground truth tracings (GTT): On the epistemic limits of machine learning. Big Data & Society, 10(1), 1–12. https://doi.org/10/gtfgvx
Peter, F. (2020). The Grounds of Political Legitimacy. Journal of the American Philosophical Association, 6(3), 372–390. https://doi.org/10/grqfhn
Robertson, S., Nguyen, T., Hu, C., Albiston, C., Nikzad, A., & Salehi, N. (2023). Expressiveness, Cost, and Collectivism: How the Design of Preference Languages Shapes Participation in Algorithmic Decision-Making. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 1–16. https://doi.org/10/gr6q2t
Rubel, A., Castro, C., & Pham, A. K. (2021). Algorithms and autonomy: The ethics of automated decision systems. Cambridge University Press.
Sloane, M., Moss, E., Awomolo, O., & Forlano, L. (2020). Participation is not a Design Fix for Machine Learning. arXiv:2007.02423 [Cs]. http://arxiv.org/abs/2007.02423
Zytko, D., J. Wisniewski, P., Guha, S., P. S. Baumer, E., & Lee, M. K. (2022). Participatory Design of AI Systems: Opportunities and Challenges Across Diverse Users, Relationships, and Application Domains. Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing Systems, 1–4. https://doi.org/10/gs8kv6

Democratizing AI Through Continuous Adaptability: The Role of DevOps

Below are the abstract and slides for my contribution to the TILTing Perspectives 2024 panel “The mutual shaping of democratic practices & AI,” moderated by Merel Noorman.

Slides

democratizing-ai-through-continuous-adaptability-the-role-of-devops-tilting-perspectives-2024 Download

Abstract

This presentation delves into democratizing artificial intelligence (AI) systems through contestability. Contestability refers to the ability of AI systems to remain open and responsive to disputes throughout their lifecycle. It approaches AI systems as arenas where groups compete for power over designs and outcomes.

Autonomy, democratic agency, legitimation

We identify contestability as a critical system quality for respecting people’s autonomy. This includes their democratic agency: their ability to legitimate policies. This includes policies enacted by AI systems.

For a decision to be legitimate, it must be democratically willed or rely on “normative authority.” The democratic pathway should be constrained by normative bounds to avoid arbitrariness. The appeal to authority should meet the “access constraint,” which ensures citizens can form beliefs about policies with a sufficient degree of agency (Peter, 2020 in Rubel et al., 2021).

Contestability is the quality that ensures mechanisms are in place for subjects to exercise their democratic agency. In the case of an appeal to normative authority, contestability mechanisms are how subjects and their representatives gain access to the information that will enable them to evaluate its justifiability. In this way, contestability satisfies the access constraint. In the case of democratic will, contestability-by-design practices are how system development is democratized. The autonomy account of legitimation adds the normative constraints that should bind this democratic pathway.

Himmelreich (2022) similarly argues that only a “thick” conception of democracy will address some of the current shortcomings of AI development. This is a pathway that not only allows for participation but also includes deliberation over justifications.

The agonistic arena

Elsewhere, we have proposed the Agonistic Arena as a metaphor for thinking about the democratization of AI systems (Alfrink et al., 2024). Contestable AI embodies the generative metaphor of the Arena. This metaphor characterizes public AI as a space where interlocutors embrace conflict as productive. Seen through the lens of the Arena, public AI problems stem from a need for opportunities for adversarial interaction between stakeholders.

This metaphorical framing suggests prescriptions to make more contentious and open to dispute the norms and procedures that shape:

AI system design decisions on a global level, and
human-AI system output decisions on a local level (i.e., individual decision outcomes), establishing new dialogical feedback loops between stakeholders that ensure continuous monitoring.

The Arena metaphor encourages a design ethos of revisability and reversibility so that AI systems embody the agonistic ideal of contingency.

Post-deployment malleability, feedback-ladenness

Unlike physical systems, AI technologies exhibit a unique malleability post-deployment.

For example, LLM chatbots optimize their performance based on a variety of feedback sources, including interactions with users, as well as feedback collected through crowd-sourced data work.

Because of this open-endedness, democratic control and oversight in the operations phase of the system’s lifecycle become a particular concern.

This is a concern because while AI systems are dynamic and feedback-laden (Gilbert et al., 2023), many of the existing oversight and control measures are static, one-off exercises that struggle to track systems as they evolve over time.

DevOps

The field of DevOps is pivotal in this context. DevOps focuses on system instrumentation for enhanced monitoring and control for continuous improvement. Typically, metrics for DevOps and their machine learning-specific MLOps offshoot emphasize technical performance and business objectives.

However, there is scope to expand these to include matters of public concern. The matters-of-concern perspective shifts the focus on issues such as fairness or discrimination, viewing them as challenges that cannot be resolved through universal methods with absolute certainty. Rather, it highlights how standards are locally negotiated within specific institutional contexts, emphasizing that such standards are never guaranteed (Lampland & Star, 2009, Geiger et al., 2023).

MLOps Metrics

In the context of machine learning systems, technical metrics focus on model accuracy. For example, a financial services company might use Area Under The Curve Receiver Operating Characteristics (AUC-ROC) to continuously monitor and maintain the performance of their fraud detection model in production.

Business metrics focus on cost-benefit analyses. For example, a bank might use a cost-benefit matrix to balance the potential revenue from approving a loan against the risk of default, ensuring that the overall profitability of their loan portfolio is optimized.

Drift

These metrics can be monitored over time to detect “drift” between a model and the world. Training sets are static. Reality is dynamic. It changes over time. Drift occurs when the nature of new input data diverges from the data a model was trained on. A change in performance metrics may be used to alert system operators, who can then investigate and decide on a course of action, e.g., retraining a model on updated data. This, in effect, creates a feedback loop between the system in use and its ongoing development.

An expansion of these practices in the interest of contestability would require:

setting different metrics,
exposing these metrics to additional audiences, and
establishing feedback loops with the processes that govern models and the systems they are embedded in.

Example 1: Camera Cars

Let’s say a city government uses a camera-equipped vehicle and a computer vision model to detect potholes in public roads. In addition to accuracy and a favorable cost-benefit ratio, citizens, and road users in particular, may care about the time between a detected pothole and its fixing. Or, they may care about the distribution of potholes across the city. Furthermore, when road maintenance appears to be degrading, this should be taken up with department leadership, the responsible alderperson, and council members.

Example 2: EV Charching

Or, let’s say the same city government uses an algorithmic system to optimize public electric vehicle (EV) charging stations for green energy use by adapting charging speeds to expected sun and wind. EV drivers may want to know how much energy has been shifted to greener time windows and its trends. Without such visibility on a system’s actual goal achievement, citizens’ ability to legitimate its use suffers. As I have already mentioned, democratic agency, when enacted via the appeal to authority, depends on access to “normative facts” that underpin policies. And finally, professed system functionality must be demonstrated as well (Raji et al., 2022).

DevOps as sociotechnical leverage point for democratizing AI

These brief examples show that the DevOps approach is a potential sociotechnical leverage point. It offers pathways for democratizing AI system design, development, and operations.

DevOps can be adapted to further contestability. It creates new channels between human and machine actors. One of DevOps’s essential activities is monitoring (Smith, 2020), which presupposes fallibility, a necessary precondition for contestability. Finally, it requires and provides infrastructure for technical flexibility so that recovery from error is low-cost and continuous improvement becomes practically feasible.

The mutual shaping of democratic practices & AI

Zooming out further, let’s reflect on this panel’s overall theme, picking out three elements: legitimation, representation of marginalized groups, and dealing with conflict and contestation after implementation and during use.

Contestability is a lever for demanding justifications from operators, which is a necessary input for legitimation by subjects (Henin & Le Métayer, 2022). Contestability frames different actors’ stances as adversarial positions on a political field rather than “equally valid” perspectives (Scott, 2023). And finally, relations, monitoring, and revisability are all ways to give voice to and enable responsiveness to contestations (Genus & Stirling, 2018).

And again, all of these things can be furthered in the post-deployment phase by adapting the DevOps lens.

Bibliography

Alfrink, K., Keller, I., Kortuem, G., & Doorn, N. (2022). Contestable AI by Design: Towards a Framework. Minds and Machines, 33(4), 613–639. https://doi.org/10/gqnjcs
Alfrink, K., Keller, I., Yurrita Semperena, M., Bulygin, D., Kortuem, G., & Doorn, N. (2024). Envisioning Contestability Loops: Evaluating the Agonistic Arena as a Generative Metaphor for Public AI. She Ji: The Journal of Design, Economics, and Innovation, 10(1), 53–93. https://doi.org/10/gtzwft
Geiger, R. S., Tandon, U., Gakhokidze, A., Song, L., & Irani, L. (2023). Making Algorithms Public: Reimagining Auditing From Matters of Fact to Matters of Concern. International Journal of Communication, 18(0), Article 0.
Genus, A., & Stirling, A. (2018). Collingridge and the dilemma of control: Towards responsible and accountable innovation. Research Policy, 47(1), 61–69. https://doi.org/10/gcs7sn
Gilbert, T. K., Lambert, N., Dean, S., Zick, T., Snoswell, A., & Mehta, S. (2023). Reward Reports for Reinforcement Learning. Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, 84–130. https://doi.org/10/gs9cnh
Henin, C., & Le Métayer, D. (2022). Beyond explainability: Justifiability and contestability of algorithmic decision systems. AI & SOCIETY, 37(4), 1397–1410. https://doi.org/10/gmg8pf
Himmelreich, J. (2022). Against “Democratizing AI.” AI & SOCIETY. https://doi.org/10/gr95d5
Lampland, M., & Star, S. L. (Eds.). (2008). Standards and Their Stories: How Quantifying, Classifying, and Formalizing Practices Shape Everyday Life (1st edition). Cornell University Press.
Peter, F. (2020). The Grounds of Political Legitimacy. Journal of the American Philosophical Association, 6(3), 372–390. https://doi.org/10/grqfhn
Raji, I. D., Kumar, I. E., Horowitz, A., & Selbst, A. (2022). The Fallacy of AI Functionality. 2022 ACM Conference on Fairness, Accountability, and Transparency, 959–972. https://doi.org/10/gqfvf5
Rubel, A., Castro, C., & Pham, A. K. (2021). Algorithms and autonomy: The ethics of automated decision systems. Cambridge University Press.
Scott, D. (2023). Diversifying the Deliberative Turn: Toward an Agonistic RRI. Science, Technology, & Human Values, 48(2), 295–318. https://doi.org/10/gpk2pr
Smith, J. D. (2020). Operations anti-patterns, DevOps solutions. Manning Publications.
Treveil, M. (2020). Introducing MLOps: How to scale machine learning in the enterprise (First edition). O’Reilly.

PhD update – January 2022

It has been three years since I last wrote an update on my PhD. I guess another post is in order.

My PhD plan was formally green-lit in October 2019. I am now over three years into this thing. There are roughly two more years left on the clock. I update my plans on a rolling basis. By my latest estimation, I should be ready to request a date for my defense in May 2023.

Of course, the pandemic forced me to adjust course. I am lucky enough not to be locked into particular methods or cases that are fundamentally incompatible with our current predicament. But still, I had to change up my methods, and reconsider the sequencing of my planned studies.

The conference paper I mentioned in the previous update, using the MX3D bridge to explore smart cities’ logic of control and cityness, was rejected by DIS. I performed a rewrite, but then came to the conclusion it was kind of a false start. These kinds of things are all in the game, of course.

The second paper I wrote uses the Transparent Charging Station to investigate how notions of transparent AI differ between experts and citizens. It was finally accepted late last year and should see publication in AI & Society soon. It is titled Tensions in Transparent Urban AI: Designing A Smart Electric Vehicle Charge Point. This piece went through multiple major revisions and was previously rejected by DIS and CHI.

A third paper, Contestable AI by Design: Towards A Framework, uses a systematic literature review of AI contestability to construct a preliminary design framework, is currently under review at a major philosophy of technology journal. Fingers crossed.

And currently, I am working on my fourth publication, tangentially titled Contestable Camera Cars: A Speculative Design Exploration of Public AI Systems Responsive to Value Change, which will be based on empirical work that uses speculative design as a way to develop guidelines and examples for the aforementioned design framework, and to investigate civil servants’ views on the pathways towards contestable AI systems in public administration.

Once that one is done, I intend to do one more study, probably looking into monitoring and traceability as potential leverage points for contestability, after which I will turn my attention to completing my thesis.

Aside from my research, in 2021 was allowed to develop and teach a master elective centered around my PhD topic, titled AI & Society. In it, students are equipped with technical knowledge of AI, and tools for thinking about AI ethics. They apply these to a design studio project focused on conceptualizing a responsible AI-enabled service that addresses a social issue the city of Amsterdam might conceivably struggle with. Students also write a brief paper reflecting on and critiquing their group design work. You can see me on Vimeo do a brief video introduction for students who are considering the course. I will be running the course again this year starting end of February.

I also mentored a number of brilliant master graduation students: Xueyao Wang (with Jacky Bourgeois as chair) Jooyoung Park, Loes Sloetjes (both with Roy Bendor as chair) and currently Fabian Geiser (with Euiyoung Kim as chair). Working with students is one of the best parts of being in academia.

All of the above would not have been possible without the great support from my supervisory team: Ianus Keller, Neelke Doorn and Gerd Kortuem. I should also give special mention to Thijs Turel at AMS Institute’s Responsible Sensing Lab, where most of my empirical work is situated.

If you want to dig a little deeper into some of this, I recently set up a website for my PhD project over at contestable.ai.

“Geen transparantie zonder tegenspraak” — betoog voor première documentaire transparante laadpaal

Het onderstaande korte betoog sprak ik uit tijdens de online premiere van de documentaire over de transparante laadpaal op donderdag 18 maart 2021.

Ik had laatst contact met een internationale “thought leader” op het gebied van “tech ethics”. Hij vertelde mij dat hij heel dankbaar is voor het bestaan van de transparante laadpaal omdat het zo’n goed voorbeeld is van hoe design kan bijdragen aan eerlijke technologie.

Dat is natuurlijk ontzettend leuk om te horen. En het past in een bredere trend in de industrie gericht op het transparant en uitlegbaar maken van algoritmes. Inmiddels is het zelfs zo ver dat wetgeving uitlegbaarheid (in sommige gevallen) verplicht stelt.

In de documentaire hoor je meerdere mensen vertellen (mijzelf inbegrepen) waarom het belangrijk is dat stedelijke algoritmes transparant zijn. Thijs benoemt heel mooi twee redenen: Enerzijds het collectieve belang om democratische controle op de ontwikkeling van stedelijke algoritmes mogelijk te maken. Anderzijds is er het individuele belang om je recht te kunnen halen als een systeem een beslissing maakt waarmee je het (om wat voor reden dan ook) niet eens bent.

En inderdaad, in beide gevallen (collectieve controle en individuele remedie) is transparantie een randvoorwaarde. Ik denk dat we met dit project een hoop problemen qua design en techniek hebben opgelost die daarbij komen kijken. Tegelijkertijd doemt er een nieuwe vraag aan de horizon op: Als we begrijpen hoe een slim systeem werkt, en we zijn het er niet mee eens, wat dan? Hoe krijg je vervolgens daadwerkelijk invloed op de werking van het systeem?

Ik denk dat we onze focus zullen moeten gaan verleggen van transparantie naar wat ik tegenspraak of in goed Engels “contestability” noem.

Ontwerpen voor tegenspraak betekent dat we na moeten gaan denken over de middelen die mensen nodig hebben voor het uitoefenen van hun recht op menselijke interventie. Ja, dit betekent dat we informatie moeten aanleveren over het hoe en waarom van individuele beslissingen. Transparantie dus. Maar het betekent ook dat we nieuwe kanalen en processen moeten inrichten waarmee mensen verzoeken kunnen indienen voor het herzien van een beslissing. We zullen na moeten gaan denken over hoe we dergelijke verzoeken beoordelen, en hoe we er voor zorgen dat het slimme systeem in kwestie “leert” van de signalen die we op deze manier oppikken uit de samenleving.

Je zou kunnen zeggen dat ontwerpen van transparantie eenrichtingsverkeer is. Informatie stroomt van de ontwikkelende partij, naar de eindgebruiker. Bij het ontwerpen voor tegenspraak gaat het om het creëren van een dialoog tussen ontwikkelaars en burgers.

Ik zeg burgers want niet alleen klassieke eindgebruikers worden geraakt door slimme systemen. Allerlei andere groepen worden ook, vaak indirect beïnvloed.

Dat is ook een nieuwe ontwerp uitdaging. Hoe ontwerp je niet alleen voor de eindgebruiker (zoals bij de transparante laadpaal de EV bestuurder) maar ook voor zogenaamde indirecte belanghebbenden, bijvoorbeeld bewoners van straten waar laadpalen geplaatst worden, die geen EV rijden, of zelfs geen auto, maar evengoed een belang hebben bij hoe stoepen en straten worden ingericht.

Deze verbreding van het blikveld betekent dat we bij het ontwerpen voor tegenspraak nóg een stap verder kunnen en zelfs moeten gaan dan het mogelijk maken van remedie bij individuele beslissingen.

Want ontwerpen voor tegenspraak bij individuele beslissingen van een reeds uitgerold systeem is noodzakelijkerwijs post-hoc en reactief, en beperkt zich tot één enkele groep belanghebbenden.

Zoals Thijs ook min of meer benoemt in de documentaire beïnvloed slimme stedelijke infrastructuur de levens van ons allemaal, en je zou kunnen zeggen dat de design en technische keuzes die bij de ontwikkeling daarvan gemaakt worden intrinsiek ook politieke keuzes zijn.

Daarom denk ik dat we er niet omheen kunnen om het proces dat ten grondslag ligt aan deze systemen zelf, ook zo in te richten dat er ruimte is voor tegenspraak. In mijn ideale wereld is de ontwikkeling van een volgende generatie slimme laadpalen daarom participatief, pluriform en inclusief, net als onze democratie dat zelf ook streeft te zijn.

Hoe we dit soort “contestable” algoritmes precies vorm moeten geven, hoe ontwerpen voor tegenspraak moeten gaan werken, is een open vraag. Maar een aantal jaren geleden wist niemand nog hoe een transparante laadpaal er uit zou moeten zien, en dat hebben we ook voor elkaar gekregen.

Update (2021-03-31 16:43): Een opname van het gehele event is nu ook beschikbaar. Het bovenstaande betoog start rond 25:14.

“Contestable Infrastructures” at Beyond Smart Cities Today

I’ll be at Beyond Smart Cities Today the next couple of days (18-19 September). Below is the abstract I submitted, plus a bibliography of some of the stuff that went into my thinking for this and related matters that I won’t have the time to get into.

In the actually existing smart city, algorithmic systems are increasingly used for the purposes of automated decision-making, including as part of public infrastructure. Algorithmic systems raise a range of ethical concerns, many of which stem from their opacity. As a result, prescriptions for improving the accountability, trustworthiness and legitimacy of algorithmic systems are often based on a transparency ideal. The thinking goes that if the functioning and ownership of an algorithmic system is made perceivable, people understand them and are in turn able to supervise them. However, there are limits to this approach. Algorithmic systems are complex and ever-changing socio-technical assemblages. Rendering them visible is not a straightforward design and engineering task. Furthermore such transparency does not necessarily lead to understanding or, crucially, the ability to act on this understanding. We believe legitimate smart public infrastructure needs to include the possibility for subjects to articulate objections to procedures and outcomes. The resulting “contestable infrastructure” would create spaces that open up the possibility for expressing conflicting views on the smart city. Our project is to explore the design implications of this line of reasoning for the physical assets that citizens encounter in the city. Because after all, these are the perceivable elements of the larger infrastructural systems that recede from view.

Alkhatib, A., & Bernstein, M. (2019). Street-Level Algorithms. 1–13. https://doi.org/10.1145/3290605.3300760
Ananny, M., & Crawford, K. (2018). Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability. New Media and Society, 20(3), 973–989. https://doi.org/10.1177/1461444816676645
Centivany, A., & Glushko, B. (2016). “Popcorn tastes good”: Participatory policymaking and Reddit’s “AMAgeddon.” Conference on Human Factors in Computing Systems – Proceedings, 1126–1137. https://doi.org/10.1145/2858036.2858516
Crawford, K. (2016). Can an Algorithm be Agonistic? Ten Scenes from Life in Calculated Publics. Science Technology and Human Values, 41(1), 77–92. https://doi.org/10.1177/0162243915589635
DiSalvo, C. (2010). Design, Democracy and Agonistic Pluralism. Proceedings of the Design Research Society Conference, 366–371.
Hildebrandt, M. (2017). Privacy As Protection of the Incomputable Self: Agonistic Machine Learning. SSRN Electronic Journal, 1–33. https://doi.org/10.2139/ssrn.3081776
Jackson, S. J., Gillespie, T., & Payette, S. (2014). The Policy Knot: Re-integrating Policy, Practice and Design. CSCW Studies of Social Computing, 588–602. https://doi.org/10.1145/2531602.2531674
Jewell, M. (2018). Contesting the decision: living in (and living with) the smart city. International Review of Law, Computers and Technology. https://doi.org/10.1080/13600869.2018.1457000
Lindblom, L. (2019). Consent, Contestability, and Unions. Business Ethics Quarterly. https://doi.org/10.1017/beq.2018.25
Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016). The ethics of algorithms: Mapping the debate. Big Data & Society, 3(2), 205395171667967. https://doi.org/10.1177/2053951716679679
Van de Poel, I. (2016). An ethical framework for evaluating experimental technology. Science and Engineering Ethics, 22(3), 667–686. https://doi.org/10.1007/s11948-015-9724-3

“Contestable Infrastructures: Designing for Dissent in Smart Public Objects” at We Make the City 2019

Thijs Turèl of AMS Institute and myself presented a version of the talk below at the Cities for Digital Rights conference on June 19 in Amsterdam during the We Make the City festival. The talk is an attempt to articulate some of the ideas we both have been developing for some time around contestability in smart public infrastructure. As always with this sort of thing, this is intended as a conversation piece so I welcome any thoughts you may have.

The basic message of the talk is that when we start to do automated decision-making in public infrastructure using algorithmic systems, we need to design for the inevitable disagreements that may arise and furthermore, we suggest there is an opportunity to focus on designing for such disagreements in the physical objects that people encounter in urban space as they make use of infrastructure.

We set the scene by showing a number of examples of smart public infrastructure. A cyclist crossing that adapts to weather conditions. If it’s raining cyclists more frequently get a green light. A pedestrian crossing in Tilburg where elderly can use their mobile to get more time to cross. And finally, the case we are involved with ourselves: smart EV charging in the city of Amsterdam, about which more later.

Image credits: Vattenfall, Fietsfan010, De Nieuwe Draai

We identify three trends in smart public infrastructure: (1) where previously algorithms were used to inform policy, now they are employed to perform automated decision-making on an individual case basis. This raises the stakes; (2) distributed ownership of these systems as the result of public-private partnerships and other complex collaboration schemes leads to unclear responsibility; and finally (3) the increasing use of machine learning leads to opaque decision-making.

These trends, and algorithmic systems more generally, raise a number of ethical concerns. They include but are not limited to: the use of inductive correlations (for example in the case of machine learning) leads to unjustified results; lack of access to and comprehension of a system’s inner workings produces opacity, which in turn leads to a lack of trust in the systems themselves and the organisations that use them; bias is introduced by a number of factors, including development team prejudices, technical flaws, bad data and unforeseen interactions with other systems; and finally the use of profiling, nudging and personalisation leads to diminished human agency. (We highly recommend the article by Mittelstadt et al. for a comprehensive overview of ethical concerns raised by algorithms.)

So for us, the question that emerges from all this is: How do we organise the supervision of smart public infrastructure in a democratic and lawful way?

There are a number of existing approaches to this question. These include legal and regulatory (e.g. the right to explanation in the GDPR); auditing (e.g. KPMG’s “AI in Control” method, BKZ’s transparantielab); procurement (e.g. open source clauses); insourcing (e.g. GOV.UK) and design and engineering (e.g. our own work on the transparent charging station).

We feel there are two important limitations with these existing approaches. The first is a focus on professionals and the second is a focus on prediction. We’ll discuss each in turn.

First of all, many solutions target a professional class, be it accountants, civil servants, supervisory boards, as well as technologists, designers and so on. But we feel there is a role for the citizen as well, because the supervision of these systems is simply too important to be left to a privileged few. This role would include identifying wrongdoing, and suggesting alternatives.

There is a tension here, which is that from the perspective of the public sector one should only ask citizens for their opinion when you have the intention and the resources to actually act on their suggestions. It can also be a challenge to identify legitimate concerns in the flood of feedback that can sometimes occur. From our point of view though, such concerns should not be used as an excuse to not engage the public. If citizen participation is considered necessary, the focus should be on freeing up resources and setting up structures that make it feasible and effective.

The second limitation is prediction. This is best illustrated with the Collinridge dilemma: in the early phases of new technology, when a technology and its social embedding are still malleable, there is uncertainty about the social effects of that technology. In later phases, social effects may be clear but then often the technology has become so well entrenched in society that it is hard to overcome negative social effects. (This summary is taken from an excellent van de Poel article on the ethics of experimental technology.)

Many solutions disregard the Collingridge dilemma and try to predict and prevent adverse effects of new systems at design-time. One example of this approach would be value-sensitive design. Our focus in stead is on use-time. Considering the fact that smart public infrastructure tends to be developed on an ongoing basis, the question becomes how to make citizens a partner in this process. And even more specifically we are interested in how this can be made part of the design of the “touchpoints” people actually encounter in the streets, as well as their backstage processes.

Why do we focus on these physical objects? Because this is where people actually meet the infrastructural systems, of which large parts recede from view. These are the places where they become aware of their presence. They are the proverbial tip of the iceberg.

The use of automated decision-making in infrastructure reduces people’s agency. For this reason, resources for agency need to be designed back into these systems. Frequently the answer to this question is premised on a transparency ideal. This may be a prerequisite for agency, but it is not sufficient. Transparency may help you become aware of what is going on, but it will not necessarily help you to act on that knowledge. This is why we propose a shift from transparency to contestability. (We can highly recommend Ananny and Crawford’s article for more on why transparency is insufficient.)

To clarify what we mean by contestability, consider the following three examples: When you see the lights on your router blink in the middle of the night when no-one in your household is using the internet you can act on this knowledge by yanking out the device’s power cord. You may never use the emergency brake in a train but its presence does give you a sense of control. And finally, the cash register receipt provides you with a view into both the procedure and the outcome of the supermarket checkout procedure and it offers a resource with which you can dispute them if something appears to be wrong.

Image credits: Aangiftedoen, source unknown for remainder

None of these examples is a perfect illustration of contestability but they hint at something more than transparency, or perhaps even something wholly separate from it. We’ve been investigating what their equivalents would be in the context of smart public infrastructure.

To illustrate this point further let us come back to the smart EV charging project we mentioned earlier. In Amsterdam, public EV charging stations are becoming “smart” which in this case means they automatically adapt the speed of charging to a number of factors. These include grid capacity, and the availability of solar energy. Additional factors can be added in future, one of which under consideration is to give priority to shared cars over privately owned cars. We are involved with an ongoing effort to consider how such charging stations can be redesigned so that people understand what’s going on behind the scenes and can act on this understanding. The motivation for this is that if not designed carefully, the opacity of smart EV charging infrastructure may be detrimental to social acceptance of the technology. (A first outcome of these efforts is the Transparent Charging Station designed by The Incredible Machine. A follow-up project is ongoing.)

Image credits: The Incredible Machine, Kars Alfrink

We have identified a number of different ways in which people may object to smart EV charging. They are listed in the table below. These types of objections can lead us to feature requirements for making the system contestable.

Because the list is preliminary, we asked the audience if they could imagine additional objections, if those examples represented new categories, and if they would require additional features for people to be able to act on them. One particularly interesting suggestion that emerged was to give local communities control over the policies enacted by the charge points in their vicinity. That’s something to further consider the implications of.

And that’s where we left it. So to summarise:

Algorithmic systems are becoming part of public infrastructure.
Smart public infrastructure raises new ethical concerns.
Many solutions to ethical concerns are premised on a transparency ideal, but do not address the issue of diminished agency.
There are different categories of objections people may have to an algorithmic system’s workings.
Making a system contestable means creating resources for people to object, opening up a space for the exploration of meaningful alternatives to its current implementation.

PhD update – January 2019

Thought I’d post a quick update on my PhD. Since my previous post almost five months have passed. I’ve been developing my plan further, for which you’ll find an updated description below. I’ve also put together my very first conference paper, co-authored with my supervisor Gerd Kortuem. It’s a case study of the MX3D smart bridge for Designing Interactive Systems 2019. We’ll see if it gets accepted. But in any case, writing something has been hugely educational. And once I finally figured out what the hell I was doing, it was sort of fun as well. Still kind of a trip to be paid to do this kind of work. Looking ahead, I am setting goals for this year and the nearer term as well. It’s all very rough still but it will likely involve research through design as a method and maybe object oriented ontology as a theory. All of which will serve to operationalise and evaluate the usefulness of the “contestability” concept in the context of smart city infrastructure. To be continued—and I welcome all your thoughts!

Designing Smart City Infrastructure for Contestability

The use of information technology in cities increasingly subjects citizens to automated data collection, algorithmic decision making and remote control of physical space. Citizens tend to find these systems and their outcomes hard to understand and predict [1]. Moreover, the opacity of smart urban systems precludes full citizenship and obstructs people’s ‘right to the city’ [2].

A commonly proposed solution is to improve citizens understanding of systems by making them more open and transparent [3]. For example, GDPR prescribes people’s right to explanation of automated decisions they have been subjected to. For another example, the city of Amsterdam offers a publicly accessible register of urban sensors, and is committed to opening up all the data they collect.

However, it is not clear that openness and transparency in and of itself will yield the desired improvements in understanding and governing of smart city infrastructures [4]. We would like to suggest that for a system to perceived as accountable, people must be able to contest its workings—from the data it collects, to the decisions it makes, all the way through to how those decisions are acted on in the world.

The leading research question for this PhD therefore is how to design smart city infrastructure—urban systems augmented with internet-connected sensing, processing and actuating capabilities—for contestability [5]: the extent to which a system supports the ability of those subjected to it to oppose its workings as wrong or mistaken.

References

Burrell, Jenna. “How the machine ‘thinks’: Understanding opacity in machine learning algorithms.” Big Data & Society 3.1 (2016): 2053951715622512.
Kitchin, Rob, Paolo Cardullo, and Cesare Di Feliciantonio. “Citizenship, Justice and the Right to the Smart City.” (2018).
Abdul, Ashraf, et al. “Trends and trajectories for explainable, accountable and intelligible systems: An hci research agenda.” Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems. ACM, 2018.
Ananny, Mike, and Kate Crawford. “Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability.” New Media & Society 20.3 (2018): 973-989.
Hirsch, Tad, et al. “Designing contestability: Interaction design, machine learning, and mental health.” Proceedings of the 2017 Conference on Designing Interactive Systems. ACM, 2017.