Public policy must regulate algorithms and AI to avoid adverse impact on societyhttps://indianexpress.com/article/opinion/columns/artificial-intelligence-algorithm-policy-5768876/

Public policy must regulate algorithms and AI to avoid adverse impact on society

A rethink of public policy is absolutely essential if non-desirable impacts of artificial intelligence on human race are to be arrested.

human brain, cloud computing network, cloud computing, neuroscience, human brain, cloud interface, artificial intelligence, b/ci, b/ci conceptmatrix style brain, neural norobotics, neurons for b/ci
Even high-profile programmers/developers may not be able to trace the working of such algorithms making nearly impossible the identification of any anti-competitive practice. (Representational image; source: Pixabay/geralt)

Businesses are increasingly utilising algorithms to improve their pricing models, enhance customer experience and optimise business processes. Governments are employing algorithms to detect crime and determine fines. Consumers are benefitting from personalised services and lower prices. However, algorithms have also raised concerns such as collusions and malfunctioning, privacy, competition issues, and information asymmetry.

Automated systems have now made it easier for firms to achieve collusive outcomes without formal agreement or human interaction, thereby signalling anti-competitive behaviour. This results in “tacit algorithmic collusion”, an outcome which is still not covered by existing competition law.

This can occur in non-oligopolistic markets too. In 2015, US Federal Trade Commission fined David Topkins (former e-commerce executive of a company selling online posters and frames), for fixing the price of certain posters sold through Amazon Marketplace using complex algorithms, impacting consumer welfare and competition adversely. In 2011, two third-party Amazon merchants, Bordeebook and Profnath, attempted to use algorithmic pricing to sell an out-of-print version of Peter Lawrence’s The Making of a Fly. The first seller algorithmically priced the book at 1.27059 times the price of the second seller. The second seller’s price was thus automatically set at 0.9983 times the price of the first seller. Over time, the price shot up to an unimaginable high of over $23 million — before dropping to $106.23 and $106.05 respectively! But, the relative pricing between the booksellers remained unchanged, indicating endless possibilities for both “collusion” and “chaos”.

Security concerns too remain paramount. In order to enjoy services at low or zero price, consumers neglect the value of their data. Access to easily procurable data such as Facebook “likes” can be used to target only advantageous customers circumventing anti-discrimination mechanisms. Application of advanced algorithms have also resulted in an increase in ransomware attacks. Ransomware is a form of malicious software that blocks access to a victim’s data and threatens to make it public unless the ransom demanded by the hacker is paid. A devastating cyber attack — the WannaCry ransomware attack — hit the world in May 2017, affecting around 2,30,000 computers across 150 countries. Through the use of EternalBlue, an exploit leaked by the Shadow Brokers hacking group, the malware spread on Microsoft Window’s Operating System across the globe without any human interaction. Such attacks require prompt action by regulatory authorities.

Advertising

Important concerns pertain to “competition” as well. Processing of large datasets through dynamic algorithms generate real-time data “feedback loops”, impacting competition adversely. As more users visit select platforms, not only more data, but data with greater reliability is collected, allowing firms to more effectively target customers. Consequently, more users feedback into this loop. Such feedback loops have the potential of creating entry barriers which are a cause of concern for competition authorities. Also, better monetisation of platforms reinforces such loops as the additional revenues generated are reinvested to improve services, thereby, further attracting more users — leading to dominance. That Google has been estimated to charge a higher cost-per-click (CPC) than Bing, a competitor, suggests that advertisers attribute a higher probability of converting a viewer of Google’s ads into a customer.

Then, we have evolving machine-learning algorithms ranging from voice recognition systems to self-driving cars. Even high-profile programmers/developers may not be able to trace the working of such algorithms making nearly impossible the identification of any anti-competitive practice. Such algorithms may have the capacity to identify a dominant strategy on their own to maximise profits.

A rethink of public policy is absolutely essential if non-desirable impacts of artificial intelligence on human race are to be arrested.

Kaur is principal, Shri Ram College of Commerce and professor of economics and public policy, University of Delhi. Sarna is assistant professor, department of commerce, SRCC