Of hype and prospects: Smart Cities, self-healing infrastructure and other myths

Recently at the Amsterdam Smart City Event 2012, we were inundated with a dizzying array of buzz-words and heady warnings concerning the need to forge ahead rapidly with analytics-driven urban infrastructure management innovations.  Compelling projections from the likes of the UN Population Division have cited sobering demographic projections:  75% of the world’s projected 10 billion population by 2050 will be living in urban zones, the majority in Dickensian conditions in exploding developing regions.  Meanwhile industrialized nation megacities are expected to continue to struggle to upgrade their own crumbling infrastructures, as highlighted in this week’s edition of The Economist, cover dated June 28th, which chronicles London’s struggle to cope with aging and outdated infrastructure against the ensuing background of the 2012 London Olympic games.

The Amsterdam Smart City Event at times devolved into a histrionic Malthusian klaxon call for radical austerity measures, lest our species drive ourselves unwittingly to extinction.  I understand keynote speaker Jeremy Rifkin is a respected author and advisor to governments and international dignitaries of all stripes.  The tone of his address was ‘fire and brimstone’ in terms of poor prospects for a stable future. His message was admirable and he offered some interesting visions extracted from his earlier works (i.e. hydrogen energy storage, buildings as energy generators). Given the sponsors for the conference, many of whom were consulting and software providers, there was often the sense that much of the rhetoric was aimed at attempting to prise open the purse strings of European city officials.

All of which returns to a theme being tracked here:  the inflating bubble associated with the hype surrounding ‘analytics’ and ‘Big Data’ as pushed by software and services marketing doyens, particularly from the ‘big brand’ firms.  The danger is that the real potential of achieving promising quantum-leaps in challenges such as urban infrastructure management efficiency might be prematurely drowned-out by carnival barker snake oil solutions and inflated dreams, all ferried forward by an unpleasant rhetorical undercurrent of millenarian doom.

Again a reoccuring theme here, there needs to be a realist-driven retrenching concerning the goals and capabilities of business analytics on two main accounts:  1) analytics starts and ends as an organizational decision-management problem (as addressed chiefly by the social and economic sciences: political science, behavioral economics, industrial sociology, etc.), and 2) the real technical advances cannot be captured or adopted as marketing sound bites – the details concerning analytics implementations are necessarily complex, involving managing complex, overlapping and at time non-linear  ‘systems of systems’.

Firstly, to the credit of the Amsterdam Smart City Event, a few bold presentations were spot-on in clarifying that no amount of analytics software or ‘gee-whiz’ initiatives such as electricity Smart Grid IT management will have a chance for evidencing substantial change without a tight engineering based on a deep understanding of the relevant human behavior factors, particularly complex multi-stakeholder dynamics and political coalition building. An earlier blog on analytics as a phenomenon resting on a foundation of organizational decision making processes is otherwise relevant to mention.

For instance, in the electricity Smart Grid case, the notion that new sensors will allow consumers to plan their own electricity purchases in a more targeted way was raised.  There seems to be an assumption that consumers will simply begin logging-in en masse to websites provided by electricity utilities.  Here they will gleeful spend hours conducting family electricity usage economic analysis to trade cents-on-the-euro (or dollar) discounts by bidding to wash their clothes at 2 am (off-peak, in the parlance).  Research has clearly proved this is a pernicious fiction:  the typical consumer is a busy and distracted agent, too busy making ends meet and, frankly living life, to engage in such wishful public austerity.

The inability of many companies to hand money to their workers via enticing ‘opt in’ pension co-contributions is a case-in-point:  what works is to make pension contributions the default (in the, perhaps patriarchal, self-interest of the worker) and to make opting-out the chosen alternative.  Such realist and common-sense understandings of how governments, municipalities, and utilities alike must ‘get real’ concerning mass human behavior is elegantly explored in Thaler and Sunstein’s recent work Nudge: Improving Decisions about Health, Wealth, and Happiness.  A well-considered key theme observed in Nudge is a nod to the ground breaking research of Daniel Kahneman concerning human decision making biases, as represented in his own work Thinking, Fast and Slow (as recently reviewed here).

Secondly, technical and methodological advances in computer-aided analytics is indeed a promising human advancement.  However, just as when software companies go off-the-rails’ as soon as the marketing department starts running the show, there is a burning need for analytics-associated data scientists, many of whom are PhD holders and have spent decades working in their field, to retain some hold on the reigns concerning attempts to interpret, communicate, achitect, and implement analytics-based solutions such as Smart Cities.  The problem of the so-called ‘Smart City’ concept is that of managing complex ‘systems of systems‘.  Many solutions offered currently, such as Smart Grid implementations, begin and end as two-dimensional software-and-sensor implementations, without considering the deeper overlaps with consumer and market social behavioral factors. I spoke directly to this topic in relation to Smart Cities at an analytics conference in London this past April:  here is the presentation…

The argument here is that advanced analytics, such as envisioned in Smart City solutions, requires hybridized techno-economic behavioral analysis, which analyzes intra-systems relationships between technology, infrastructure, markets (i.e. commodity price dynamics), consumer behavior, and political stakeholder interests.  This is a multi-disciplinary challenge that can indeed be addressed with emerging advanced analytics methods and tools (i.e. predictive analytics mixed with simulation mixed with systems dynamics analysis mixed with stakeholder multi-criteria analysis).  However, to the degree marketers simply sell a software implementation, the results are going to be disappointing at best, and potentially destructive at worst.

In closing, I will pose a brief example of the proper treatment of an advanced analytics solution domain, that of  ‘self-healing infrastructure’ (a term which is at times posed in the context of Smart Cities).  This is a compelling buzz-term:  it sounds quite lovely to the ears indeed – self-healing!  However, this is when the ‘marketing hype’ warning bells should go off and where the real experts should step in to have entrenched conversations concerning the fundamental technical prospects.  If marketers sell such solutions and the result is simply a license for expensive anlaytics software, we are indeed all doomed in the Rifkin sense!

Let me give an example of a real technical prospect for ‘self-healing infrastructure’ associated with analytics in an attempt to illustrate that this should not be driven purely by warm feelings and sound bites.  The prospect of immense ‘Big Data’ datasets now emerging as a result of sophisticated digital sensor arrays associated with complex infrastructure (i.e. oil pipelines, municipal water management, gas turbines, public transport, electricity grids, etc.) is giving rise to the possibility of machine-driven learning for optimization and pro-active maintenance.

Here are two very promising techniques in closing, teasers which will be expanded upon in a future article:  1)  automated cluster analysis to produce inferential understanding (extrapolation) of infrastructure failure conditions in order to target proactive improvements to infrastructure, and 2) the combination of sophisticated analysis methods to produce powerful new approaches, such as machine learning decision tree extraction combined with Real Options Analysis (ROA) in order to automate maintenance and improvement decision making in complex infrastructure.

These are not techniques that should (or could) be designed by marketers and sold to stakeholders.  Rather, there needs to be careful discussion among data scientists, staged testing, and pilots driven by stakeholder collaborations.  More on this to come…