The Future of Data Science in the Age of COVID-19

According to McKinsey⁴, the rout of 'face-to-face' society has broken historical analytics models so badly that pre-COVID data may itself have to be considered in a separate historical context, and compensated for in any attempt to maintain a continuum of measurement for year-on-year progression in business cycles.

At the time of writing, post-outbreak data has had less than a year of existence. With no way to know when the post-COVID era will begin (or whether the threat will diminish with the speed of the Millennium Bug or the languorous, decades-long struggle against HIV), it's difficult to understand at what point a long-term analytical context will emerge that isn't specifically related to efforts against the virus.

Even if we were inclined to consider the period since the outbreak as a 'hard reset' in analytics and forecasting, not only is post-outbreak data insufficient yet to form a base for prediction, but it's also defined by a number of pivotal factors that are in no way constant or reliable enough for forecasting further years of combatting the epidemic. These factors include:

• State-led fiscal interventions that give a statistically skewed representation of national and global consumption, productivity and stability⁵.
• Supply chain interruptions, which have had a significant effect on downstream economic logistics management⁶.
• Changes in customer behavior that relate directly to the increased short-term and long-term economic insecurity, and which are not in themselves constant or predictable⁷.
• The ambient effect⁸ of the diminution, destruction and/or bankruptcy of multi-billion dollar industries that underpinned the pre-COVID model, such as airlines⁹ and hospitality¹⁰. 

There is general agreement, however, on one upcoming scenario in which the data science sector will not only play an important role, but need to negotiate in itself.

As occurred after the 2008 financial crisis, economically resistant global tech giants are hoovering up the exodus of data scientists from commercial startups unable to weather the storm¹⁸. 

Although these well-funded entities — including Google, Amazon, Microsoft, and Apple — are in themselves powerful economic engines for data science and machine learning research, and although all of them are major contributors to open-source AI that has multiple applications, their common interests center on analytics: data gathering, modeling, prediction and deployment, primarily around user intent and with an emphasis on natural language processing (NLP) and computer vision.

It seems likely that the ongoing monetization of analytics and forecasting systems will represent the central artery keeping more abstract or experimental sub-sectors alive on a provisional basis in the profit-driven context of the private sector, as the global economy struggles to weather the fiscal effects of the pandemic.

Ongoing gains and losses are sector-dependent: according to one industry report¹⁹, the healthcare analytics industry is predictably set to grow at the highest CAGR in the period 2020-21, while the general analytics market is forecast to expand by 15.5% in response to the tectonic cultural changes of 2020.

There is a strong possibility that the customary recovery environment of a late-stage recession will be different from those of the past; and that the inability to quickly reboot into a pre-COVID economic template will elevate data analytics to a more central and enduring position³⁰ in civic and commercial culture, no matter what the outcome of the crisis might be.

If we do not currently need the new roads, train lines, and other travel-based infrastructure boosts that kick-started the failing economies of the more mobile societies of the past^31,32, and if we are unable to re-capitalize (as after 2008) by continuing to inflate the cost of real estate in major cities in the face of an urban exodus³³ and an emerging long-term telecommuting culture³⁴, it seems inevitable that money will follow consumers into virtual space — which signifies a massive acceleration of digital conversion and the analytics culture that accompanies it.

For business, this external pressure to migrate from analogue transactions to digital paradigms and platforms is merely an acceleration of what was already occurring in the years leading to 2020.

The following are the key areas where data science is poised to aid in sector recovery and endurance.

The use of AI to model historical and 'live' logistics data allows distribution companies to simulate thousands of scenarios³⁵, incorporating ad-hoc factors such as state and national restrictions that may come into force or be cancelled according to infection levels and local and national policies.

As with many sectors that depend on predictive machine learning systems, the Supply Chain Risk Management (SCRM) industry has had to recover from unexpected variations in demand since the late winter of 2020, and is in some cases merging prediction models for the lifetime of the COVID-19 crisis into traditional logistics forecasting systems³⁶.

In the United States alone, the coronavirus crisis is set to force a 733% increase in the number of Americans working full-time from home³⁷, with many influential companies now committed to partial or permanent long-term telecommuting³⁸, irrespective of societal progress against COVID-19.

Though there are diverse AI-powered solutions to aid remote workers in scheduling, search, and even to improve the quality of teleconferencing³⁹, it seems likely that innovations in AI-based remote monitoring will also bring new data streams that can be utilized for performance monitoring at an individual and team level⁴⁰.

The use of data science and machine learning to evaluate employee performance pre-dates the pandemic⁴¹. Such systems are centralized and can be applied to the remote working sphere via authorized VPNs, with only minor latency considerations compared to on-premises workstations.

One company has experimented with AI-enabled fitness monitors⁴² to assess the level of employee stress during remote working under COVID-19. However, depending on the geographic region and its applicable laws, the proliferation of such systems may eventually bring privacy and diverse legal implications⁴³ into the public arena, notwithstanding an employee's contractual consent.

Machine learning has been used in a number of evaluative systems for students forced to follow academic courses in ad-hoc online environments^44,45 — a necessary innovation that may eventually feed into more stable and generalized online management systems.

Besides providing curated information on the COVID-19 crisis⁴⁶ for overburdened health services, chatbots are proving an essential stopgap⁴⁷ for companies whose former investment in intelligent virtual assistants (IVAs) has become urgent under coronavirus-related restrictions, and for whom negative public perception⁴⁸ of AI-based assistants has been relegated by the scale of the current public health crisis⁴⁹.

Thus it appears that increased public tolerance for chatbot-based interaction in core services such as health may be improving the general reputation and uptake of chatbot development on the whole.

The pre-COVID estimate that chatbot sales would grow to US$112 billion by 2023 is now thought by some to have multiplied significantly⁵⁰. However, the current trend towards economic decline also indicates that lower general economic demand could cancel out this trend, leaving the market with similar prospects to those forecast in the pre-COVID era⁵¹.

The necessity for increased engagement with chatbots represents an unforeseen opportunity to improve both their performance and reputation: as the volume of IVA transactions increases, analysis of the interactions allows their underlying systems to improve in a manner that could not have been foreseen prior to the pandemic — a further boon to chatbot research and deployment.

According to Deloitte⁵², the possibility of ongoing disruption to supply chains, together with diminished spending patterns, constitutes a threat to the hardware technology release cycle, including computing components, smartphones, consumer electronics devices, and a range of business technologies.

This situation could trigger a short- or long-term emphasis on after-market support, where the role of machine learning analytics could transform from a minor industry advantage to a critical edge⁵³ in a market often beset by low margins.

The much-criticized tendency of manufacturers to abandon software support for older models in order to perpetuate the buying cycle could even transform from time-limited after-market care to the subscription-based SaaS models that have come to dominate the software market since 2015⁵⁴.

However, as with many COVID-led contingencies, this kind of short-term model revision could make it difficult to eventually return to the more profitable annual hardware churn if the crisis continues long enough to normalize the increased product retention and extended after-market maintenance — with all but the most resilient and consolidated tech manufacturers facing massive losses and an existential threat⁵⁵.

With an emerging shift away from cities⁵⁶, subsequent turbulence in the rental and real estate market^57,58, and the sporadic obstacles in visiting potential acquisitions in person, the use of AI in real estate seems set to grow.

Prior to COVID-19, machine learning was already engaged in lead generation, mortgage lending risk assessment⁵⁹, budgetary management⁶⁰, property value prediction⁶¹ and conveyancing⁶². Since the outbreak, diverse platforms and products have arisen that offer interactive virtual viewings, with AI-enabled augmented reality apps posited for massive and unforeseen growth in this and other sectors⁶³.