Infosys’ Responsible AI Framework and Data Analytics Drive Ethical Innovation and Sustainability for Enterprises

CXOToday has engaged in an exclusive interview with Sunil Senan, Senior Vice President and Business Head, Data and Analytics, Infosys

Q1. How do you navigate the challenge of balancing the benefits and risks associated with big data and AI technologies from an ethical standpoint?

Even as AI creates breakthrough opportunities, it is opening up questions around bias, security, privacy, and trust. Infosys has developed and adopted a “Responsible AI Framework” to overcome ethical AI challenges and build trustworthy systems powered by Infosys Topaz. This framework addresses five dimensions – People & Planet, Economic Context, Data & Input, AI Model, and Task & Output. Each of them further has its own sub-dimensions, e.g., People & Planet has stakeholders, human rights, etc. Within Data & Input, there is the critical sub-dimension of “Managing Bias”, which includes processes and audits to ensure that demographic diversity is represented in training data and that systemic injustices to disadvantaged populations are filtered out.  For every prospective AI project, information is collected along these dimensions and sub-dimensions that address issues including the tenets of ethical AI such as fairness, transparency, accountability, privacy, and security. Application of this framework results in the computation of a “Risk Score” for the project that identifies its risk category. If the risk category is among acceptable ones, the project is approved to go ahead. Otherwise, steps for further risk mitigation are recommended or the project is rejected.

Q2. The pace of technological advancement often outpaces regulatory efforts. How do you see the role of regulation in addressing the ethical implications of big data and AI? Are there any gaps that need to be addressed.

Just as there is the phenomenon of “moral hazard” in economics, wherein actors have the incentive to make risky investments when they do not bear the full cost of their actions (i.e., relying on government bailouts), there is the emergent force we call “learning hazard”.  This phrase encompasses the idea that enterprises (or practitioners) are incentivized to train their models on as much data as possible, however, they don’t realize the risks and costs they accumulate in the process in terms of data ownership issues, biases in the data, IP ownership of learnt models or meeting ethical standards, etc. This includes collecting and using data whose provenance is not 100% established for model training purposes {e.g., creative commons content, or GPL (General Public License), an open-source software}.  It might also include a collection of personal data without compensation, infringement of proprietary corporate know-how, or other violations of protected intellectual property.  Regulations might be slow to address these gaps, but strong practices to manage and document the lineage of knowledge for training generative-ai models can avoid the risk of being blindsided by regulations when they do become law.

Q3. In the context of user data, how does Infosys ensure that informed consent is obtained and that user data is used responsibly? Can you share any specific use cases or mechanisms Infosys employs?

Infosys practices principles of privacy by design to ensure transparency, minimal data collection & consented processing of personal data for all the stakeholders in our ecosystem including customers, employees, and vendors, among others. We provide apt notices to the users at the time of data collection & provide mechanisms so that users can take an informed decision to opt-in or opt-out, be it a physical or online/digital consent, revocable at any time. Our solutions/applications are responsible by design.

Infosys makes every effort to protect the personal information that comes under its purview with its Personal Information Management System (PIMS). PIMS ensures user data is used responsibly as it consists of time-tested data processing practices such as purpose limitation, maintaining records, fine-grained access controls, etc. Our PIMS is the convergence of international best practices, client-prescribed requirements, and applicable data privacy regulations across geographies. We are among the first few organizations globally, to have our PIMS certified with accreditation to the ISO 27701 privacy information management standard. We also conduct regular audits to ensure that user consent is explicitly obtained & enforced.

Infosys Topaz delivers these best practices to our clients and helps them build privacy practices to sustain compliance in the ever-changing data privacy regulatory landscape across the globe. For example, for an American multichannel video programming distributor, we helped them define and implement the Privacy by Design process along with necessary controls to ensure data privacy compliance throughout the software development process. This included consent collection and proper processing of data based on the available consent.

Q4. What steps does Infosys take to identify and mitigate bias in data sets and algorithms?

Bias in AI can emanate from multiple ways: bias in the training datasets/historical data being trained, people designing the AI/ML models, and people interpreting the model outputs. As an organization, we are continuously evolving and learning from our research and experiences and are applying these learnings to our “Responsible by Design” practices.  Not only do we strive for accuracy and relevance in our AI deployments, but also design to the highest possible ethical and socially responsible standards.  We mitigate bias in our training data and models using well-defined practices such as

• Diversify & pre-process training data
• Monitor and track bias in model outputs
• Provide @prompt engineering training
• Apply bias-mitigation practices, including the application of techniques such as adversarial training, counterfactual data augmentation, re‑sampling, etc.
• Create a culture of transparency and openness, encouraging user feedback

Our “Responsible by design” framework, as part of Infosys Topaz, enables us to regularly evaluate and audit AI systems to identify and rectify potential biases.

Q5. How do you foster a culture of privacy and ethics within your organization, encouraging employees to prioritize these aspects in their work with big data and AI? Especially with a premium offering like Topaz, is any upskilling or reskilling of the workforce required?

Infosys proactively takes a design-led approach to every customer engagement. We keep users’ experience at the forefront of everything we do. We then mold technology to fit the desired experience rather than doing it the other way. This approach ensures that we are prioritizing secure & ethical practices while implementing any solution.

Infosys has a dedicated DPO (Data Privacy Office) team in place strongly backed by the senior management to foster a privacy-first culture. We conduct mandatory data privacy, security & ethics training regularly for each employee to ensure that employees are aware of the best practices & responsibilities while handling data. The training content is updated regularly to keep up with the latest technological developments & risks. We also provide regular guidelines, conduct drills, and organize engaging events with lucrative incentives for the employees to participate in & become an advocate of privacy & ethics themselves. For example, we are among the first few organizations globally to have our PIMS certified with accreditation to the ISO 27701 privacy information management standard.

With the rapid advancement of technology & a premium offering such as Topaz, it is imperative that the workforce is also trained to foresee & handle the challenges & risks associated with these technologies. The workforce creating, using & getting impacted by these technologies are trained to make sure both our internal and external offerings are ethical, user-centric, transparent & accountable. This approach goes beyond mere compliance with regulations & standards, emphasizing proactive & ethical decision-making that considers the broader implication of design choices.

At Infosys, some aspects in which reskilling/upskilling of the workforces are being done revolves around:

• AI literacy
• Ethical AI design & principles such as understanding bias, fairness, transparency, accountability, data privacy, etc., so that these are integrated into the development of AI technologies.
• User research methodologies & user-centric design approaches to gather insights, empathize with users & integrate their needs & values into the design process.
• Collaborate across disciplines such as bringing together AI experts, designers, ethicists & domain specialists.
• Data & AI governance & privacy practices to handle sensitive user data in AI systems.
• Responsible decision-making frameworks & ethical reasoning to help the workforce assess & navigate complex ethical dilemmas associated with AI technologies.

We have curated these capabilities as part of Infosys Topaz to help our customers inculcate this culture through our state-of-the-art offerings and solutions. Our revamped offerings ensure a smooth transition of our customers’ workforce for a secure & ethical implementation of Infosys Topaz.

Q6. Businesses are dedicating considerable resources to achieving sustainability in their operations, and Infosys has consistently kept up with all the latest sustainability practices. How does Infosys empower enterprises to embrace sustainability by harnessing the power of data analytics?

Companies are collecting and sharing a growing amount of sustainability data. However, businesses often fail to take advantage of the benefits that their data can provide. This critical cycle of collection and sharing can lead to insights that both improve ESG outcomes and help the enterprise achieve its business goals. ESG data and its disclosure can help companies meet investor expectations, comply with regulations, improve branding, and build operational efficiency. Thus, Data-Powered Sustainability Transforms ESG from an Obligation to a Business Opportunity. Infosys Topaz is helping leading Companies to meet their ESG needs through our Infosys Topaz AI-powered data tools, accurate intelligence (data and insights), a common reference architecture, and data & AI economy play with an ESG-specific partner ecosystem integrated into ESG Intelligence Cloud.

This addresses Observability – Capturing data at every touchpoint to measure and respond – closed loop, Ecosystem Integration – enabling the seamless flow of information from upstream providers and through the different stages of the production process to downstream consumers, Governance, Validation & Certification across the Value Chain and Sustainability Collaboration and Exchange

Infosys has deep experience in enabling clients to achieve practical outcomes from sustainability initiatives through our Sustainability Intelligence cloud.

• For example, for a global bank, we helped them focus on both measuring financed emissions for regulatory disclosures as well as assessing the risk and transition impact from climate change programs to ensure portfolio optimization & meeting global climate commitments.
• For a leading waste processor, we enabled a sustainability data hub to deliver data products through an information marketplace and provide a ‘complete’ view to customers to connect their Scope 3 goals with their waste processing focus, a key component of end-of-life carbon impact and emissions.
• For an apparel major, we are helping them meet their 2025 goals of achieving 100% traceability and 100% sustainable-source materials for their products by creating a digital twin of their products throughout the product lifecycle with visibility up to the original lot of the materials sourced.