It's the dream of big data and digital transformation... sophisticated tools that provide you with insightful strategic, commercial and real-time information for your organisation. The benefits of algorithms in business are obvious: dynamic pricing algorithms that maximise revenue and create efficiencies, rapid data-supported decision making across the business, agile predicting of market trends, greater customer satisfaction through competitive pricing, more tailored products and services.
It's no wonder more and more businesses are depending on algorithms and machine learning for success. But what's less obvious is where ethical and compliance lines are drawn. When does competitive advantage become a harmful impact – what are the rules around this and who is responsible?
More than ever, boards and management need to know what their algorithms know, what they are doing and what they might do, in order to navigate the new frontier between compliance and competitive advantage.
Algorithmic business – competing in the age of AI
Algorithms are a set of instructions that enable a computer to pull together and process different information sources to produce a result. Once limited to the largest corporations, they are now part of everyday life – from adjusting product prices in real-time and making personalised shopping suggestions to matching people on dating websites.
Both buyer and seller may benefit from:
- pricing algorithms that predict the optimal profit maximising price for individuals (personalised prices) or groups of individuals
- search algorithms that order results to most appeal to the individuals
- clustering algorithms that seek to group consumers
Algorithms can have huge strategic and efficiency benefits, facilitating revenue growth and cost reduction. Well-designed algorithms allow firms to more effectively target consumers and to operate more efficiently by faster adjustments to prevailing market conditions.
Pricing algorithms, for example, allow repricing in real-time or may engage in price discrimination providing better-individualised quotes (eg, through using credit checks). Consumer costs can be reduced, and meaningful relationships facilitated between buyers and sellers or those within a group.
Algorithmic business is changing the competitive landscape – dynamic pricing mechanisms are just one example of how they provide an advantage to those who use them. They also provide an excellent illustration of how automated algorithms can go rogue.
When algorithms go wrong
More sophisticated algorithms can learn, including from other algorithms, and can create their own instructions. For example, pricing algorithms may use similar sources of inputs, including competitor prices, to produce a price which then becomes an input into the competitor's pricing algorithm. The faster and more accurate price adjustments have benefits through better matching of supply and demand.
However, this might also lead to harm as these algorithms may learn to tacitly collude and reduce the benefits of price competition. Writing in the Harvard Business Review, law professor Ariel Erzachi and Maurice Stucke identified three ways in which pricing algorithms may lead to tacit collusion:
- Hub and spoke – competitors use the same algorithm with the same input data, and these become a hub to coordinate prices (akin to horizontal collusion)
- Predictable agent – each firm unilaterally develops an algorithm that reacts to market events in a predictable way, allowing competing firms to capture signals and leading to coordinated outcomes
- Artificial intelligence – sophisticated pricing algorithms self-learn and anticipate market events, potentially leading to coordinated outcomes
Hub and spoke cartels and collusion risk
In its 2018 paper on pricing algorithms, the UK's Competition and Markets Authority (CMA) said hub and spoke was the most concerning because it simply requires firms to adopt the use of the same algorithm. This could mean that, if the online retailers purchased the third-party pricing algorithm knowing that their competitors had also done so, they could potentially be considered to knowingly be engaging in collusive behaviour.
Algorithms may also lead to consumer harm through reduced or manipulated choice, for example, through manipulation of choice architecture, algorithmic discrimination and the creation of barriers. They may also lead to dark patterns – subtle interfaces for tricking the user – making it difficult for consumers to transparently compare prices, for example.
Who's responsible for the machine?
Algorithms are increasingly under the spotlight of competition authorities. In 2015 David Topkins, founder of online poster retailer Poster Revolution, was the first e-commerce business to be prosecuted under antitrust law by the US Department of Justice. The adoption of specific pricing algorithms that collected competitors' pricing information was considered to be coordinating changes to pricing strategies for the sale of posters on Amazon Marketplace – essentially creating a digital cartel.
In this case, the use of algorithms to help execute the digital poster cartel's task had the same effect as a cartel executed by humans. The European Commissioner for Competition, Margrethe Vestager, summed it up as "companies can't escape responsibility by hiding behind a computer program".
The general principle under EU law is that companies will be held liable for any anti-competitive practices of their employees, even if they can show that they have used their best efforts to prevent such behaviour. Vestager adds: "What businesses need to know is that when they decide to use an automated system, they will be held responsible for what it does. So they had better know how that system works."
Company algorithms – actions for the board
The CMA Data, Technology and Analytics (DaTA) team has been focusing on algorithms, with a new Digital Markets Unit (DMU) being established to implement a pro-competitive regime for digital markets and to develop an oversight regime for algorithms. All boards will be responsible for understanding what their organisation's algorithms are doing – and for ensuring that they comply with the new regime and do not inadvertently breach competition law. In a CMA research paper published last year on how algorithms can harm consumers, the message was clear that "firms are responsible for effective oversight of such systems, which should include robust governance, holistic impact assessments, monitoring and evaluation."
The challenge for those responsible is that it is not always clear how business algorithms are creating benefits – or harm. It may also be difficult to follow how the algorithms are learning and the impact of their often automated actions. And while it may be possible to code an algorithm not to fix prices, in the age of AI, it may be that a self-learning algorithm learns that pricing coordination gives the most profitable outcome.
Identifying tacit collusion versus a competitive response to demand and supply fluctuations is tricky. When you don't know if collusion is occurring, how can you comply with CMA requirements?
Three actions for boards to consider now
- Compliance – make sure you understand the changing policies (regulations) around algorithms and your responsibilities to avoid inadvertently breaching competition law and/or falling under the scrutiny of the CMA
- Be aware – understand how your algorithms are working now and what they may do in the future
- Know the risk – take a risk-based approach to the development, adopting and adaptation of algorithms and embed it into your governance processes
To find out more about how we can support you, please get in touch with Schellion Horn.
