IEEE Technology and Society Magazine - March 2021 - 57

Another critical factor is the lack of collaboration among administrations, the so-called smart
collaboration [11]. There is little incentive to collaborate because administrations tend to perceive
their needs differently and because local politics
prioritizes serving and siding with the agenda of the
local constituency over promoting new projects and
innovation. This stands in dramatic contrast to the
highly dynamic and collaborative world of AI, which
is dominated by Open Source software. This field
operates on collaboration between industry and
academia, nonprofits such as Apache or NumFocus and industry giants such as Google, Facebook,
or Microsoft. Moreover, we cannot forget the series
of visible and invisible barriers surrounding public
administrations: difficulties to attract and retain talent within government jobs because salary levels
are much lower than in the private sector; lack of
diversity within internal teams or among established
service providers-both key enabling elements for
moving past the single solution to the problem.
Furthermore, venture capitalists and start-ups
generally see the public sector as very complicated
in terms of value capture. It is not enough to present
a proposal that solves the problem and attracts the
attention of the public, but the added requirement to
solve the market problem and sell it to a myriad of
public organizations full of bureaucratic procedures
and vested political interests. This situation discourages most of the start-up ecosystem from engaging
with the public sector.
Finally, one more obstacle to the use of disruptive
technologies in cities' administrations is the risk and
uncertainty associated with it. In many areas, the
reward is not clear, while the costs and the implementation effort needed is high. Due to their public
accountability, cities tend to be cautious about the
use of resources in areas of high uncertainty like this
one. Therefore, the use of external grants to fund
pilots for exploring new areas is ideal for helping cities in blue sky exploration.
The long-term success of AI in the public sector
depends on the integration of these technologies
into services that allow the implementation of policies that benefit citizens and the administration.
However, governments cannot carry out this integration by themselves; they need the collaboration
of public universities, in the public and private
sectors. These cross-sectoral partnerships involve

March 2021

complex management challenges that hinder their
success [12].
The same AI technologies that help to improve
the functioning of services and decisions in the private sector can contribute to solving some of the
challenges that stifle public administrations such as
reducing waiting times in the provision of services,
detecting fraud, reducing the management burden, improving decision-making, or freeing up time
for critical tasks. We portray some use cases of AI
applied in different public sectors.
Health care: The multiple applications of AI in the
health care are well known, for example, the use of
deep learning (DL) techniques and computer vision
in clinical imaging [13], demonstrating promising
results in diagnosis spanning radiology [14], [15],
ophthalmology [16], [17], or pathology [18], [19].
There are other applications of AI in the field of
health care related with administrative aspects. For
example, in the optimization of health inspections in
restaurants using social media data to reduce food
poisoning and derived hospital admissions [20], outpatient visit forecasting in hospitals [21] or optimization of bed rotation in hospitals [22]. The potential
of AI in the field of public health care still has many
frontiers to be explored, but we must be aware that
medical data are among the most sensitive personal
data and that with undesired processing they can
produce serious harm to citizens [1].
Public safety: AI has started having a significant
impact on the law enforcement agencies (LEAs)
community and is being promoted through numerous initiatives funded by the European Commission:
from autonomous drones to patrol and detect forest wildfires [23], detection and counter-measuring
criminal unmanned aerial vehicles [24], using facial
recognition to locate wanted criminals [25] or natural language processing (NLP) in social media to
detect radicalization [26], among others. AI techniques are also used for cybersecurity, for example,
in malware detection [27], intrusion detection, and
prevention systems optimization [28]. An example
of AI put at service of LEAs was demonstrated by a
public data aggregation technology that could pinpoint in 22 s the position of the shooter in Las Vegas
Mandalay Bay tragedy by processing social media
[29]. By using AI, LEAs will be able to respond in
a more efficient way to incidents, prevent threats,
improve resource utilization, and analyze criminal
activity in a new dimension that was unachievable

57
IEEE Technology and Society Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - March 2021
