Cyprus Center for Algorithmic Transparency

Project Information

CyCAT

Grant agreement ID: 810105

DOI

10.3030/810105

Project closed

EC signature date 8 June 2018

Start date 1 October 2018

End date 31 December 2021

Funded under

• Twinning of research institutions

Total cost

€ 999 965,00

EU contribution

€ 999 965,00

Coordinated by

ANOIKTO PANEPISTIMIO KYPROU (OPEN UNIVERSITY OF CYPRUS)
Cyprus

Periodic Reporting for period 2 - CyCAT (Cyprus Center for Algorithmic Transparency)

Reporting period: 2020-01-01 to 2021-12-31

Summary of the context and overall objectives of the project

Despite having one of the highest rates of tertiary educational attainment in Europe, multiple indicators suggest that Cyprus is at risk of falling behind in the Fourth Industrial Revolution. For instance, on the Digital Economy and Society Index, Cyprus continues to rank 21st or lower within the EU, during the past five years (2017 – 2021). Human Capital is consistently cited as the key dimension of concern. These findings are worrying considering the rapid changes taking place in the information landscape, such as the consolidation of networked information services, and the rise of proprietary algorithmic processes that mediate citizens’ access to information and opportunities. Furthermore, the workforce of the future will need to develop a healthy symbiosis with algorithmic processes, driving the need not only for digital skills, but also algorithmic literacy.

The Cyprus Center for Algorithmic Transparency (CyCAT), established at the Open University of Cyprus (OUC), aims to become the regional expert in issues of information access, working with local authorities and educators to promote transparency in algorithmic systems, and enhanced digital literacy skills. The CyCAT Twinning project established a network of leading researchers across Europe and Israel, who are experts in the areas of informatics and algorithmic bias, enhancing the capacities of the local CyCAT team.

Specifically, the project’s objectives focus on making both scientific and broader, societal impacts:

• Scientific: Provide support to OUC researchers, enhancing their relevance and performance, by establishing a strong network and developing outward-facing, cross-institutional initiatives. Through CyCAT scientific activities, the team will develop novel computational methods, personalized tools, and innovative training and education programs.
• Societal: Promote digital literacy and raise citizens’ awareness of the ways in which data analytics and algorithmic processes directly or indirectly affect them; support the Responsible Research and Innovation culture in Cyprus.

The conclusions of the action can be summarized as follows:

• Scientific Objectives: i) The OUC significantly raised its capacity for research excellence, as evidenced by the number of competitive research proposals submitted and won during the 39 months of the project, the high-profile publications that were produced, as well as the recognition of its researchers through invitations to give seminars and keynotes, a best demo award received at an international conference, and other indicators.
• Societal Objectives: Educational interventions took place during the project, focused on three target groups – public school teachers, students in technical programs (e.g. computer science, data science), and the general public. Multiple appearances in the Cypriot media raised the issue of integrating algorithmic literacy into educational. Notably, a short (8 weeks) online course on “Everyday AI” was launched at the OUC; 396 members of the public participated.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

WP1: Project management
The end-of-project event was held under the auspices of the Deputy Ministry of Research, Innovation and Digital Policy of the Republic of Cyprus. Stakeholders from the government sector, the industry, and the media attended.

WP2: Dissemination and outreach
The public was engaged through several channels (Web/social media, traditional media, Researchers’ Night, OUC Everyday AI course, educational videos). The scientific community was engaged via the organization of a Winter School, as well as workshops. The team published results in high-profile international conferences and journals. Results were used to develop two courses. One is a self-guided module for students and researchers. The second is an eight-weeks’ long course entitled “Everyday AI,” which assumes no technical background.

WP3: Understanding social and cultural consequences of algorithms
A literature survey was carried out to understand approaches to promoting Fairness, Accountability, Transparency and Ethics (FATE) in algorithmic systems. This revealed several findings: i) solutions are scattered across research communities; ii) solutions focus on pre-processing, in-processing, and post-processing interventions; iii) there are multiple ways to define concepts such as “fairness” or “transparency”; iv) it is possible that a system can pass an audit process as being “bias free,” yet it is not perceived as being fair by its users. Our survey paper has been accepted for publication in the prestigious journal, ACM Computing Surveys.

WP4: Promoting algorithmic transparency
Based on the outputs of WP3, WP4 developed a holistic conceptual model for promoting FATE. Multiple stakeholders were identified and their roles in FATE processes better defined. Furthermore, easy-to-read guides were developed for specific groups (e.g. developers, users, but also educators / teachers), laying the groundwork for WP5.

WP5: Designing and evaluating interventions
Educational interventions were designed for two groups: i) a three-hour teacher training, in collaboration with the Cyprus Pedagogical Institute; ii) a ten-hour developers’ seminar was delivered to computer science students at the University of Cyprus. A tool-based intervention focused on helping users understand bias in a Google News Search for articles related to COVID-19. The tool was developed through participatory workshops with participants.

WP6: Inter-institutional networking
A Dagstuhl seminar “Transparency by Design,” took place in June 2021. Visits through a short-term staff exchange program, enabled researchers to collaborate on focused mini-projects. Grant-writing workshop sessions were held. Finally, the Algorithm Watchdog activity explored the potential of establishing an independent body to provide oversight to AI applications by involving citizens.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

We pursued three types of interventions to address algorithmic bias – research, educational and tool-based interventions.

• Research. Our systematic literature review on the mitigation of algorithmic bias, revealed that while the research on FATE has flourished, the limitations of disciplinary approaches, which are strictly technical or educational, are clear. To be trustworthy, AI applications must be technically robust, but aligned with human values and laws; human oversight of the behaviors of AI is needed. However, people’s algorithmic literacy must be upscaled if they are to participate in such oversight.
• Educational. Multiple initiatives were conducted, targeting three audiences: i) public school teachers, ii) those with a technical background iii) the public.
• Tool-based. The COVID news bias prototype helps users explore bias in Google News articles related to “coronavirus” and other similar terms. OpenTag is a platform through which users can explore social bias in computer vision algorithms. The tools are now being used by the team for both research and educational/demonstration purposes.

Both scientific and societal impacts have resulted from the implementation of the project. On the scientific side, CyCAT has raised the S&T capacity of the OUC team. On the societal side, CyCAT will continue to offer services that focus on raising the public’s AI literacy, promoting better, healthier relationships with AI technologies.

CyCAT Project Logo