This is the seventh theme set out by the Network, and is led by : Ewa Luger at the University of Edinburgh. It addresses the following questions:
The application of data-driven systems within social and economic spheres “requires the transformation of social problems into technical problems” (Crawford & Whittaker, 2016: 19). This translation is not direct, as it requires some reframing of the problem so that it can be articulated within the constraints of what we might design. Consent, for example, is easily designed into a system as a set of information followed by a check box, but we know that this mechanism does not tell us whether the consenting individual was capable of consenting, whether their choice was freely made, or whether they fully understood the implications of sharing their data.
More generally, contemporary data-driven systems are largely not designed within notions of human agency, data legibility or redress at the forefront. This means that there exist no ready grounds upon which any algorithmic determination, or perceived harm, might be contested by those impacted. An inability to meaningfully contest decisions made by an AI-driven system could have the effect of reinforcing existing power asymmetries or unintentionally creating new ones. There are also concerns that the more dominant an organisation, the more able it is to (re)define what constitutes ethical practice. So, how might we reveal power asymmetries by design?
Equally, unlike other technological development (e.g. the Internet, which allowed users to be curators and creators of content), AI is undemocratic in its design, in that the power to develop and train an AI lies only in the hands of organisations that have the datasets, computational power and specialised skillset required to develop such systems. Whilst there is no doubt that in many cases there is a desire to create value-neutral systems, the reality is that such systems both embed values, and fail to reflect social diversity, and thereby are not designed to meet the goals of a plural society or global community. So, how can we ensure that the values we enshrine in our systems are balanced, and how can we make visible the inequities that limit human agency?
Projects funded by the call
Because of lockdown, no workshop took place for this call. However the call was very successful, and eventually funded four projects, though their setup and progression was constrained by the ongoing situation.
Infrastructural Interactions
Helen Pritchard, University of Plymouth
How can data infrastructures contribute to the flourishing of public life? Infrastructural Interactions is a transdisciplinary research initiative inquiring into the different ways that harmful data practices extend into public services and in response seeks to develop and imagine technical alternatives to Software-as-a-Service and agile solutions.
Data-infrastructures, made up of cloud and mobile computational infrastructures, are concentrated in the hands of a few companies. They purposefully promote data intensive services running on these infrastructures, rather than pre-packaged and locally run software instances. These data- infrastructures range from health databases, border informatics, data storage warehouses, to city- dashboards for monitoring citizen flows, educational platforms and the optimisation of logistics. Data infrastructures generate harms and damage beyond ethical issues of privacy, ownership and confidentiality, through the depletion of resources for public life.
To address this we will investigate how data-infrastructures capture public data by interfacing (public) institutions and their constituents through Software-as-a-Service solutions, therefore reconfiguring their public mandate and narrowing their modes of functioning to forms of logistics (delivering a “solution” to a “need”), and optimization.
Drawing on trans*feminist, queer and anti-colonial perspectives, the research project develops tools for political and creative agency for situations perceived as in the public interest but outside of public intervention. We will work with activists (specifically groups organising with refugees and around Anti-Racist, Trans, Queer and Sex work), artists and technologists to ask how they are mobilising creative forms of organising and inventive methods to ensure that data can support their practices instead of extracting from it. How are these practices not just responding but also proposing new modes of imagining technology in the public interest? What can creative practice bring to understanding data-infrastructures and their alternatives? And how effective are these creative responses or new infrastructures?
Infrastructural Interactions is initiated by an interdisciplinary team of researchers with widely recognized experience in technological analysis, community involvement, activist practice and cultural development. They have previously collaborated on technological practice from feminist, queer and anti-colonial perspectives. The project will consist of 8-10 interviews, 2 workshops, a digitally mediated event, a research paper and a digital workbook.
Understanding the selection and aggregation of telemetric data for sousveillance practices
Sandy Gould, University of Birmingham
Employers increasingly leverage technology to collect data on their employees’ behaviour. Workers have little say in which kinds of data collected, aggregated and stored or how these data are used in decision making. Meanwhile, workers lack recourse to clear measures of work such as those found in the scientifically-managed factory, depriving them of a common quanti- tative basis upon which to organise and collectively bargain around pay, productivity, time, and conditions. We will build on our nascent collaboration combining Human-Computer Inter- action research (Gould, Cecchinato) and the Sociology of Work (Pitts). Seed funding from the ESRC Productivity Insights Network has allowed us to develop a cross-disciplinary perspec- tive on workplace tracking, telemetry and data collection (Cecchinato et al., Accepted). We propose that negotiated collective agreements could allow workers greater, more meaningful agency over the data collected about them. There is a need for an exploratory empirical investi- gation of ideas around collective negotiation of (and through) data to establish whether the ap- proach could curb the worst excesses of workplace surveillance and provide a way for data to inform claims made by workers in collective bargaining and negotiation around the value, time, productivity and conditions of their work
In this project we will build an empirical base by conducting interviews to understand how (if at all) workers are involved in decisions by their employers about data pertaining to worker ac- tivities is collected, stored, aggregated and used, and for what purpose(s) (objective 1). We will use then use these interview findings to develop prototype protocols and techniques that would help workers understand this data (and other data that might potentially be collected about them) and to critically identify different ways of understanding the data collected about them and their work. The prototypes that will enable this exploration will take the form of scenarios, personas, and schematics of potential negotiation frameworks. These aim to increase the legi- bility of data such that workers can better influence the collection and use of data and use the “evidence” which that data provides about worker activity and working environments in order to negotiate more effectively (objective 2). We will then take these prototypes into focus groups to determine whether they exhibit the properties required to help workers understand the data that is and could be collected about them and their work and to agree collective control of data around which to organise and bargain (objective 3).
Conducting this empirical work will allow us to determine whether collective agreement about what is tracked and how it is tracked in workplaces is an idea that might be deployed practi- cally in workplace bargaining and negotiation. If there are indications that it can be, then we will use our results as the basis of a larger, more comprehensive research programme partner- ing workers, employers and trade unions to develop collectively negotiated tracking and data collection. In this way, the collection of workplace data can be put on a firmer ethical footing, enhancing agency, legibility and negotiability of workplace data for workers.
Zoom Obscura: creative interventions for a data ethics of video conferencing beyond encryption
Pip Thornton, University of Edinburgh, et al.
Zoom Obscura builds on investigators’ existing and developing work into various privacy, security and ethics concerns raised by the power and influence of technology platforms in today’s increasingly datafied society and economy (Speed 2018; Thornton 2018; Dwyer 2020; Elsden 2020; Thornton, Dwyer, Duggan & Elsden – forthcoming). Accelerated by the COVID-19 pandemic, video-calling platforms such as Zoom, Microsoft Teams and Google Hangouts have become a normalized means of communication for both work and leisure. Due to the speed and scale of the crisis, there has been little time for the ethical implications of the processing and potential exploitation of the personal and biometric data accessible to these platforms to be critically assessed. Debates about data privacy and security are often framed around technological solutions such as encryption, however this project aims to approach the problem from a more human-centred ethical approach, using playful and creative intervention to allow users of these technologies to regain some political and ethical agency in their increasingly normalized roles as subjects of the camera’s gaze.
The project aims to deconstruct/unpack the Zoom assemblage, bringing together a group of artists/hackers/creative technologists with diverse and unique practices and skills in a three-part workshop series to brainstorm and prototype methods of obfuscation, subversion and other ways in which users of these technologies might regain agency in how their data is represented and how they can participate in online spaces on their own terms. Interventions might include cheap analogue hacks such as placing stickers onto webcams with watermarks, for example, or digital hacks such as adding copyrighted brands to video streams, pixilating faces, or augmenting the stream with different methods, or indeed inventing alternatives to Zoom. Interventions might lie in the hands of the users, or in the camera, the browser, the WIFI, or even in filtering the data and permissions granted to the Zoom app itself. Framed by 3 workshops covering i) ideation, ii) prototyping and iii) presentation of various interventions, the project brings together scholars from Digital Art, Cybersecurity, HCI and Digital Humanities as well as experts in commercial and cultural data ethics and a local digital arts and activism collective. In addition to the completion of the stated aims and objectives, Zoom Obscura also raises the following research questions which will inform academic outputs and further the portfolio of research of which this project is a part:
- RQ1. What are the ethical, privacy and security implications of the newly ubiquitous use of video calling platforms such as Zoom since the COVID-19 pandemic?
- RQ2. How can we better gain agency and legibility over the value potentially generated from the facial, voice, written and other data generated by the increased use of such platforms?
- RQ3. What role can critical and creative design play in challenging the accepted hegemony of video calling by reclaiming agency over the terms on which we participate in such spaces and the data we wish to present?
- RQ4. How can issues around privacy, security and agency be addressed beyond the narrative of encryption?
Human Motion Analysis – Agency, Negotiation and Legibility in Data Handling
Fani Deligianni, University of Glasgow
Empowering patients to have an active role in managing their conditions is a significant objec- tive in a growing e-Health (digital Health) era. Countries recognise this and there is strong initi- ative within Europe and beyond to transform privacy data laws and regulations to allow AI ad- vances to provide valuable insight in the field of healthcare informatics without compromising users’ privacy. The objective is to build home patient-centric systems that are intuitive to un- derstand and transparent to allow controlling the information passed to them. However, trans- lating recent advances in wearable technology and computer vision in home care is challeng- ing for two reasons. Data privacy and ethics should be encoded in the algorithms early in the pipeline so the systems are resilient to attacks and do not compromise real-time interaction. Furthermore, data quality would be significantly lower than in current experiments that take place in controlled environments with specialised personnel and advanced equipment. To en- hance trustworthiness and confidence in the technology robust designs and evaluation should be developed.
In this project, we suggest the development of online 3D platforms that firstly would enable self-monitoring of human pose, secondly, they will allow patients to interact with doctors and physiotherapists to get feedback about their progress and thirdly AI systems would provide real-time guidance and highlight errors in their posture and exercise routine. The key point is that this functionality will be enabled via AI technology that disentangles human pose, activity and biometrics via end-to-end deep learning. In this way, interpretable representation of pa- tients identity can be identified and filter out early along the processing pipeline. Emphasis will be given to protect the privacy of subjects not involved in the experiment but they share the same physical space with the patients as part of their family and friends. Therefore, their data would be captured automatically even though they have not consent, which is a scenario that poses major ethical concerns that are typically overlooked in human-pose estimation. Finally, intuitive interactive designs will be developed to allow the patient to visualise the information processed and retained by the system. Furthermore, the users would be able to control any information transmitted online to doctors and physiotherapist in a transparent way. We expect this approach to have several applications that range from e-health for elderly and patients with neurologic and mental health conditions as well as sports training and entertainment.
Human Data Interaction 