About a month ago, IT major Wipro made the headlines for offering its latest tool ETHICA to its clients. Its proprietary platform Holmes is already a market leader in offering a range of cognitive services, and the company harnesses cognitive computing, cloud capabilities, analytics to help clients retain a competitive edge in a super digital world. According to reports, this decision to roll out Ethica is part of the CEO Thierry Delaporte’s larger aim of ramping up digital technology solutions.

No matter the reason, this is a welcome move for India’s IT sector – which has notably moved away from the moniker of being the leader in Business Process Outsourcing to one that's an innovation hub for hardware and software products. Wipro has been one of the earliest architects of this success, and Holmes is one of the main drivers of this success. Holmes has been deployed in a range of information-driven verticals including BFSI, retail, manufacturing and telecom.

The introduction of a specific tool like ETHICA, which stands for Explainability, Transparency, Human-first, Interpretability, Common sense, and Auditability, is another sign of the changing times, and how organisations like Wipro are spearheading these winds of change for India as a leader in AI. Ethical AI is one of the hottest topics in governance and policy today worldwide. It is a critical component to retain and strengthen consumer trust as well. While algorithms are not inherently biased, data and training of models could lead to the introduction of biases, which could have a detrimental effect on an organisation’s credibility to service clients.

Research by IDC indicates that by 2021, algorithm opacity, decision bias, malicious use of AI, and data regulations will result in the doubling of spending on relevant governance and compliance staff and explainability teams. By 2020, 35 US states and five non-European countries will have passed GDPR-like laws, making privacy a global requirement and driving growth in outsourced privacy risk and third-party data services.

Over the years, as AI’s adoption has risen, it has also sparked several debates surrounding its lack of neutrality and perpetrating human biases. If a decade ago, the concerns were over AI making workforces redundant, today’s argument is about AI’s responsibility towards society. While governments across the world have already begun their research into developing guidelines for the responsible use of AI, organisations need to move at the same pace, if not faster. The need for responsible AI is not just being voiced by consumers and the media, but policy makers as well.

Wipro’s decision to offer its Ethica tool to clients opens a plethora of opportunities for other software majors to follow suit. Staying true to the intent of ethical AI, this will encourage enterprises leading digital transformation to offer their solutions to make AI more reliable, explainable, ethical and bias-free. 

With the premise that humans will always remain responsible, and the organization is a key partner in that responsibility, ETHICA prioritises customer trust above all else.

Here are some broad thoughts on how Wipro’s most important stakeholders will benefit from Ethica:

Concealing certain types of data from a learning model could alleviate downstream bias. This is relevant in the banking sector. If a consumer is applying for a loan, his name, PAN card, gender etc are needed for identification, but having these data points for loan processing could lead to awry results. Masking data from this level would be useful.

Deploying transparency and explainability is another important area, which is relevant to banking. While doing a Know Your Customer (KYC), using easily explainable data parameters like payment and purchase history would be more accurate in helping onboard a customer, instead of the usual data like gender, background and race, which further the scope for bias.

Anomaly detection - anomalies are sought out by developers (not necessarily a rule-based engine) in the form of fraud/duplication regardless of the background or history, like in insurance fraud. These anomalies are based on historical data without biases rather than focusing on the person or cause of the fraud, which leads to biases.

Revenue forecasting without bias - This involves the prediction of a company’s revenues by analysing multiple parameters without biases.

Human-based auditing – Every time a critical action is taken, a human is in the loop, acting as a filter for bias.

Stories: Freepik