participate in, contest or otherwise challenge
their decision-making (in terms of inputs, logic
or outcomes). Consequently, there is likely to
be power asymmetry, for instance, between
the AI system developers, service providers or
third parties, and those who interact with the AI
systems. Moreover, AI systems may fail to give a
comprehensible explanation of their underlying
decision-making process to the affected
individuals. This opacity and power asymmetry
not only expand opportunities for potential
exploitation, but may erode the sociotechnical
foundations of justice, morality and human
rights (Yeung, 2018). Some researchers, such as
Baú and Calandro (2019), have recommended
a human rights-based approach to digital
technology. Furthermore, if the decision-making
76
]
DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION
process of the digital system is hidden from the
person directly affected by the outcomes, then
the person may not trust the system. This is why,
for instance, research for explainable AI is being
conducted (see, for example, Rudin, 2019).
Currently, there are few FSN efforts aimed
at increasing the interpretability and
explainability of AI systems. Khan and Hoffmann
(2003), for instance, propose and describe
a menu construction using an incremental
knowledge acquisition system (MIKAS). The
diet recommendation system asks experts
to provide an explanation for each of their
actions, to include the explanation in the
system’s knowledge base. Interpretability
and explainability of algorithms ought to be
prioritised beyond performance and error
rates (Côté and Lamarche, 2021). Algorithm
developers, model builders and domain experts
could provide explanations for the application’s
decisions for inclusion in the system’s
knowledge base and output. Open-source
initiatives can also contribute to interpretability,
transparency and explainability of systems. For
instance, the details of a model can be fully
described within source code. It is, however, also
important to be aware that information other
than the source code may be required to fully
understand a model, including the nature of
data, documentation, etc. (Sampson
et al., 2019).
Digital technologies that are transparent and
give users freedom of choice are desirable.
In order to mitigate the risks associated with
digital technologies, it is also valuable to build
the capacity of users. For instance: providing
users with full information, including on risks
and biases; educating users about their digital
rights and responsibilities; ensuring that users
are trained or supported to handle relevant
technologies; creating an enabling environment
for users to access the required digital
infrastructure and digital resources; etc. It is
important to include stakeholders in the needs
analysis, design, piloting and implementation
of digital technologies. When users are involved
in the process, they are more likely to provide
contributions to the system development process
and trust and accept the realized systems
(Maguire, 2001).
Another concern associated with digital
technologies is who owns the FSN digital data,
who has access to it, and who has control over
its use and implementation. Issues of ownership,
access to and control of data can lead to risks
associated with inequitable data access, power
asymmetry, negative exclusive property regimes
over data, exclusion (wilful or not) of certain
types of data, unethical tracking and targeting
(for instance, through AI-powered unethical
target advertising), and market dominance by
organizations and bodies that control the data
(
SEE BOX 29
). In the process, digital technologies
could affect the cultural fabric and identity of
FSN stakeholders (Klerkx, Jakku and Labarthe,
2019) – for instance, what it means to be a
farmer (Burton, Peoples and Cooper, 2012;
Carolan, 2017), and a possible change in the
culture of farming from a hands-on approach to
data-driven management (Butler and Holloway,
2016; Carolan, 2017). Moreover, there are
cyber-security risks associated with digital
technologies in FSN (for instance, for smart
farming as described by Barreto and Amaral
[2018]). Users and respondents in FSN may be
concerned about the privacy, protection and
misuse of their data. They may fear that their
data may be used to exploit them, may be used
against them, may end up in the wrong hands,
or may put them in precarious positions in the
future. Some researchers (e.g., Clapp and Ruder,
2020) have argued that digital technologies can
reinforce existing systems which are considered
economically, socially and ecologically
unsustainable and favour specific FSN players
(Rijswijk et al., 2021).
[
77
4
NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA
BOX 29:
CHALLENGES WITH DIGITALIZING SERVICES AND ACCESS: THE CASE OF INDIA’S AADHAAR
IDENTIFICATION NUMBER
India’s Aadhaar (literally “the foundation” in Hindi) programme, intended to provide a unique 12-digit identification
number to 1.3 billion Indian residents, was launched in 2009 as a voluntary biometric ID system to smooth delivery of
public services, such as food assistance and welfare benefits, and reduce fraud. However, since 2014, the biometric
ID system under Aadhaar is being made compulsory to access more and more basic services and entitlements.
Failure to obtain an Aadhaar number has sometimes hindered residents’ access to fundamental benefits such as
rice or wheat at subsidized prices, an important source of food security for many Indians, access to pensions, school
admissions for children and so on, and this is why proper functioning of the system is essential. Implementing this
system presents a number of challenges that have led to shortcomings. Some of these shortcomings are caused
by limited availability of the necessary IT infrastructure, including electricity, to operate the biometric ID systems,
especially in rural areas. In addition, if someone is unable to go in person for biometric identification, the benefits
cannot be accessed. While delegation systems exist on paper, in practice they rarely work. This disproportionately
affects the elderly and the disabled, and those from remote villages. Furthermore, repeated reports of data leaks
have raised concerns for the privacy of personal records, which is particularly worrying as the Aadhaar identification
number is not only a condition to receive social support, but increasingly linked to private transactions, including tax
payments.
The issues presented in the implementation of the Aadhaar programme should be used as a learning experience to
exercise caution in the adoption of new digital technologies when these are linked to fundamental access to food and
social protection, as possible technology, infrastructure and capacity constraints can deeply affect the realization of
the right to food for the neediest and exacerbate inequalities.
Source: Khera, R. (2019)
In response to the risks associated with data
ownership, access and control, a responsible
research and innovation (RRI) approach to
digital transformation has been proposed for
use, for instance, in agriculture (Barrett and
Rose, 2022). The RRI approach is based on
four main principles: anticipation, inclusion,
responsiveness and reflexivity. Similarly, Rose
and Chilvers (2018) propose a more systemic
approach to map innovations associated with
digitalisation in agriculture; broadening of
notions of inclusion in RRI to include a diversity
of stakeholders; and evaluating responsible
innovation frameworks in practice to determine if
innovation processes can be made more socially
responsible.
It is also important to formulate and enact
laws, regulations and policies on ethics,
consent, privacy, data protection, ownership,
fair competition and copyright. Governments
and regional and international organizations
should involve stakeholders in defining and
implementing appropriate data standards and
policies in order to minimize the potentially
negative consequences of data access and
sharing. Ge and Bogaardt (2015) studied a
number of data harvesting initiatives in agrifood
chains to identify the key governance issues
to be addressed. Examples of data protection
and privacy laws and regulations include the
European Union’s General Data Protection
Regulation (
https://gdpr-info.eu/
) and the Data
Protection Act of the United Kingdom (of Great
Britain and Northern Ireland) (
https://www.
legislation.gov.uk/ukpga/2018/12/contents/
enacted
). Such laws and regulations are often
subject to the oversight of an independent
authority to ensure compliance and protection
of individual rights. At a broader level, the UN
Global Pulse has developed Privacy Principles
78
]
DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION
in consultation with experts from various
sectors. The United Nations Secretary-General’s
Independent Expert Advisory Group on a Data
Revolution for Sustainable Development has
recommended the development of a global
consensus on principles and standards
concerning legal, technical, privacy, geospatial
and statistical standards to, among other things,
facilitate openness and information exchange
and promote and protect human rights (FAO,
2017; UN, 2015) It is worth noting the UN High-
Level Committee on Programmes (HLCP) is
looking into a global data governance framework
(
https://unsceb.org/session-report-369
).
It is particularly important for FSN actors
to protect potentially vulnerable segments
of society. For instance, Kraak
et al. (2020)
propose various actions to protect young
people from irresponsible digital marketing
that could negatively impact diets and lifestyle
choices. Among these proposed actions are
recommendations that technology firms develop
policies to protect the digital privacy rights of
young people; enforce standards for digital
platforms that support responsible marketing
to children and adolescents; and ensure that
digital marketing and media policies are posted
on the firms’ public websites. Kraak
et al.
(2020) also propose that governments develop
comprehensive national legislation, regulations
and policies that protect digital privacy and
restrict the use of all forms of digital marketing
to children and adolescents; collaborate with
international and regional bodies to develop
cross-border policies to regulate transnational
digital marketing and media practices; monitor
and evaluate how transnational companies are
using digital marketing and social media and
enhance accountability for their practices. (More
details on governance of FSN data are presented
in Chapter 5).
It cannot be overstated that early and continuous
inclusion and involvement of all relevant
stakeholders is key to the acceptance and
success of new technologies in the FSN sector.
Stakeholders include, but are not limited to,
governments, industry, consumer groups, NGOs,
farmers and other smallholder producers.
Although upstream and downstream sectors
influence the adoption of technologies by
farmers, they can also learn from farmers so
that the technologies implemented take into
account the requirements of the farmers (OECD,
2001). In order to ensure that everyone is in a
position to benefit from new technologies and
that technology-based efforts do not reinforce
the digital divide, it is important to ensure that
digital technology implementations are adapted
to the needs, requirements and contexts of all
users and stakeholders, especially vulnerable
groups and those in developing countries who
have less digital access (due to low internet
connectivity, for example) and human capital
(for instance, related to low literacy levels).
Of course, support to ensure access to and
utilisation of technologies should indeed be
provided for all stakeholders, especially for those
who are vulnerable. Furthermore, during the
conceptualisation, design and implementation
process of such efforts, it is also important
to take into account indirect and long-term
effects of the digital technologies. Moreover, it is
instructive to create spaces for FSN stakeholders
to reflect on how digitalization will affect existing
FSN innovation systems (Bronson, 2019; Klerkx,
Jakku and Labarthe, 2019) and to consider a
policy-driven strategic overview of FSN needs
and priorities (Regan, 2021).
Involving users and stakeholders in the design
and implementation of digital applications early
and throughout the process, it becomes possible
to anticipate and address the associated risks
and needs (Rijswijk
et al., 2021) and significantly
increases the likelihood that they will accept,
value, own, support and trust the respective
technologies. Ortiz-Crespo
et al. (2021) describe
a user-centred design process that was used
to develop a system called Ushauri, to provide
farmers in Tanzania with agricultural advice.
Furthermore, with the required capacity
(such as skills, infrastructure such as open-
source tools), local individuals and groups can
themselves build digital technology platforms.
In fact, Carolan (2022) argues that participation
or inclusivity extends beyond simply making
sure that voices are heard and that, inclusivity
includes empowering individuals to build their
[
79
4
NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA
own digital platforms – in contrast to exclusive
intellectual property regimes.
If experts, users and stakeholders are not
involved in the design, development and
implementation of digital technologies, other
risks may arise, for instance in data collection,
analysis and interpretation. For example, in the
case of automated analysis or if there is a lack of
analytical expertise, without the participation of
experts and/or local stakeholders there is a risk
of misinterpretation or overgeneralization. (This
risk arises, for example, when the computational
models/algorithms used in the technologies
do not take into account the social, economic,
cultural and natural complexities of the target
people or country). Although machine learning
can help to improve prediction in nutrition-
related research (for instance in cardiovascular
risk prediction [Rigdon and Basu, 2019]),
procedures for model validation in nutrition
research are often not sound or not well reported
(Christodoulou
et al., 2019), hampering an
objective model comparison in real world case
studies. Methodologies for model development
and validation should therefore be more
carefully designed and reported (Christodoulou
et al., 2019); or improved upon (Espel-Huynh
et al., 2021). In this regard, experts can play
a key role in identifying algorithms that have
optimal performance and are appropriate for
specific prediction problems. When designing,
developing, implementing and researching
digital technologies, therefore, it is important
to involve relevant experts to give inputs for or
during the various data cycle stages (including
data collection, building the underlying models
and performing analysis).
Digital technologies should offer FSN services
and FSN content that are based on and adapted
from trusted sources, and that take into
consideration local contexts in order to meet the
unique needs and preferences of different user
groups (FAO, 2013b). For equity and inclusivity,
the global community and international
organizations should actively and continually
engage and support the sustainability of
indigenous knowledge, innovations and capacity
from the grassroots and local levels, and
vulnerable groups, thus contributing to empower
local communities.
On a final note, it is interesting to note that some
of the new digital technologies, when made
accessible, can be used to support and enhance
stakeholder engagement, promote inclusion,
and support coordination in FSN efforts. These
facilitative technologies include crowdsourcing,
crowdsensing, online social media and mobile
computing.
QUALITY OF DATA
Data quality entails elements such as accuracy,
completeness, timeliness, validity and
consistency. While digital technologies can
enhance data quality (for instance, by validating
accuracy and ensuring timeliness), there is also
potential for digital technologies to affect data
quality negatively. Data collection from users
or respondents through technologies such as
online social media, crowdsourcing and other
mobile computing-based applications is relatively
subjective and, therefore, subject to factors such
as deception and carelessness. It has also been
reported that data collected from
citizen science
efforts tends to be noisy, that is unreadable by
analysis programmes (Kelling
et al., 2015). It
may be useful to complement such user-focused
digital technologies with other digital technologies
or methods that are more objective.
It is worth noting, however, that over-reliance
on numeric data (on the false presumption
that such data is more objective) may lead to
a scenario where data or information remain
largely incomplete. In many cases, qualitative
data captures key information about local
contexts where FSN interventions are or will be
undertaken in ways that cannot be represented
by numbers. As was noted in Section 4.1.2, some
of the new and emerging digital technologies
support the processing of qualitative data in
the form of images, videos, audio recordings
and text. Such technologies can be used
for data collection (e.g. online social media,
crowdsourcing and other mobile computing-
based applications), data storage (e.g. cloud
computing, big data), data analysis (e.g. machine
80
]
DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION
learning via sentiment analysis) and data
dissemination (e.g. information visualization,
online social media). However, a number of
digital technologies still only collect or process
numeric data, given that qualitative data
collection, processing, codification and storage
may involve complex processes and be highly
demanding in terms of time and resources. Still,
it is important to consider that over-reliance
on digital technologies that collect, or process
only numeric data may downplay important
nuances that can be gleaned from qualitative
data, and therefore, it is useful to also use
digital technologies that can effectively manage
qualitative data.
Moreover, another potential source of inaccurate
data can derive from distractions in the
respondents’ uncontrolled settings, which can
affect the quality of data collected. IoT and
sensors can give false or misleading readings
(for instance due to environmental complexities),
which can translate into potentially detrimental
agricultural and nutritional decisions and
actions by farmers and policy-makers. However,
research (such as Hariri, Fredericks and
Bowers, 2019), is in progress to overcome these
limitations, making big “poor-quality” data
more valuable than small “high-quality” data.
As such, digital technologies used in real-world
settings should be constantly monitored, tested,
calibrated and enhanced and, in some cases, a
combination of digital technologies or methods
should be used to ensure data quality.
INTEROPERABILITY OF DATA
Interoperability makes it possible for different
systems to share, exchange and understand data.
This is critical when efforts are being made to
integrate different systems, which, in turn, is key
to making digital technologies and systems widely
useful. Interoperability may be necessary at any
stage in the data cycle. For instance: users may
want their respective digital applications to be able
to fetch and analyse FSN data from diverse big
data or
cloud computing
sources.
Interoperability initiatives often involve tasks
such as developing standards and specifications
(such as using ontologies to provide global
term identifiers and frameworks to define
and categorize FSN-relevant terms); building
mappings for many different sets of standards
and specifications; curating multiple domains of
vocabulary; and harmonizing existing vocabularies
or curating new terms in them. It is important to
note that initiatives, such as FAO’s AGROVOC,
17
FoodOn (Dooley
et al., 2018) and CGIAR’s Crop
Ontology,
18
(described earlier), are efforts that
contribute to interoperability.
CAPACITY, EQUITY, SCALABILITY AND
SUSTAINABILITY
Digital technologies involve relatively high
investment costs, and are expensive for
some organizations, for farmers with lower
socioeconomic status, and for other vulnerable
FSN stakeholders. Some organizations that carry
out data collection and analysis are finding the
cost of requisite technological infrastructure
prohibitive (Sivarajah
et al., 2017), in addition to
lacking sufficient personnel with skills in core
data competencies (for example, in data analysis,
information visualization, interpretation and
decision making). Vulnerable FSN stakeholders
might also not have the capacity to use the
technologies or interpret data results, or may
lack altogether access to internet connection and
digital devices. The use of digital technologies in
such scenarios may lead to or reinforce existing
inequalities, such as the digital divide, and to the
unequal distribution of the benefits of new digital
technologies, favouring those who can already
afford them. Furthermore, if technologies are
implemented without the inputs of vulnerable
FSN stakeholders, they may become even more
disconnected from and further marginalised.
It is therefore important to invest in the necessary
technology, infrastructure and research
necessary to improve data interoperability and
data quality, as well as access to and affordability
17 For further information, see
https://www.fao.org/agrovoc/about
.
18 For further information, see
https://bigdata.cgiar.org/digital-
intervention/crop-ontology-2/
.
[
81
4
NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA
of technology. It is also important to build and
enhance human capacity. For instance, by
training in core data competencies (e.g., data
collection, data analysis, information visualization,
interpretation and decision making). Several
institutions are supporting this training in the FSN
domain. For example, FAO is offering training in
such areas through the FAO eLearning Academy
(Remotely Sensed Information for Crop Monitoring
and Food Security - Techniques and methods for
arid and semi-arid areas (
https://elearning.fao.
org/course/view.php?id=155
). Other means for
building and enhancing human capacity include
educating users to support the data lifecycle
process; enhancing user- and indigenous-capacity
to improve data quality; and educating data
owners and data producers about privacy, consent,
data usage, data ownership and the rights they
have. All stakeholders – data owners, producers
and respondents – should be informed about
the purpose of collecting, processing and using
data and whether the data will be shared with
other parties. Collaboration can also be useful in
addressing some capacity concerns. The potential
benefits of collaboration include: ensuring
interoperability of technology standards and
architectures; defining appropriate data standards
and policies on data access and data sharing;
pooling digital resources and infrastructure;
implementing shared services in a synergic
manner; sharing best practices and mutually
beneficial information; developing context-relevant
and user-relevant technological interventions;
and limiting the potential for technology to be a
disincentive to meaningful production. In addition,
efforts on interoperability of data and systems
can lead to the realisation of open-source tools
and materials which, in turn, can reduce some
capacity costs. Responsible automation, as
described earlier, can also help to alleviate some
of the capacity challenges.
Furthermore, almost all digital agriculture
initiatives have to contend with the challenges
of scaling (how to include locations, users, etc.)
and sustainability (how the initiatives can extend
beyond the current funding, etc.) (Florey, Hellin
and Balié, 2020; Kos and Kloppenburg, 2019).
Some of the recommendations to overcome these
challenges include: demonstration of the benefits
of using digital technologies and tools to support
decision-making, adoption of interdisciplinary
approaches and interconnectedness, recognizing
the need for learning, feedback, partnerships, and
joint action in multi-stakeholder settings within
the context of FSN innovation systems (Florey,
Hellin and Balié, 2020 p.135; Schut
et al., 2016;
Shepherd
et al.,
2020).
82
]
Chapter 5
INSTITUTIONS
AND GOVERNANCE
FOR FSN DATA
COLLECTION,
ANALYSIS, AND USE
82
]
Italy, 15 October 2018, FAO Headquarters – CFS annual Plenary.
©IFAD/Daniele Bianchi
[
83
5
INSTITUTIONS AND GOVERNANCE FOR FSN DATA COLLECTION, ANALYSIS, AND USE
P
revious chapters in the report have made
the case for the importance of using data
to inform decisions; discussed the type
of data needed at various levels in the wide
ecosystem that determines food security and
nutrition; commented on their current availability
and the most important gaps; presented
examples of valuable initiatives that contribute
at each step of the cycle; presented an overview
of the major constraints and bottlenecks that
still affect FSN data systems worldwide; and
introduced the enormous potential that resides
in new and emerging digital technologies.
One theme that emerges from this discussion is the
increased complexity of modern data systems, with
many actors involved. Nowadays, in every practical
context – including FSN policymaking at global,
national or local levels – the collection, processing
and use of data to reach effective, evidence-
informed decisions involves a distributed (and often
fragmented) process, with responsibilities held by
different individuals and institutions, at different
levels. Ensuring the proper coordination and
collaboration among the various actors involved
throughout the data cycle is fundamental to the
success of any solution. This presents significant
challenges for the design of an effective data
governance system and creates an opportunity to
take a systems approach not only to describe what
is meant by food security and nutrition (Clapp
et
al., 2021; HLPE, 2020), but also when addressing
the roles that data collection, dissemination
and analysis play in ensuring food security and
adequate nutrition for all.
The multiplicity of actors involved in generating
and using data for public good, together with
the special nature of data in the digital era,
creates complex challenges for data governance.
This is a very active area of investigation, and a
consolidated view of which the most appropriate
governance mechanisms are and which
institutions should lead and coordinate them is
still far from being crystallized. In fact, there is
not even an agreed-upon of data governance.
The DAMA Guide to The Data Management
Body of Knowledge defines it as, “The exercise
of authority and control (planning, monitoring,
and enforcement) over the management of
data assets.” (DAMA International, 2009, p.37).
Abraham, Schneider and vom Brocke define it
as “a cross-functional framework for managing
data as a strategic enterprise asset”, highlighting
its broad scope to include the specification
of “decision rights and accountabilities for
an organization’s decision-making about its
data” and the formalization of “data policies,
standards, and procedures” (Abraham, Schneider
and vom Brocke, 2019, p.425-26). These
definitions refer to data as an asset owned by a
specific firm, company or organization (reflected
implicitly in the expression “decision-making
about
its data”), that has clearly established,
full authority and control over the data. We find
these definitions too narrow to be applied to
84
]
DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION
FSN data, which encompasses a wide range
of types of data, including data “owned” by
governments, data “owned” by private entities
and, quite importantly, data apparently owned by
no one, which is potentially available to anyone
who has the skills to access it from the internet.
The evolving data landscape that is taking shape
as the digital revolution continues, especially
following recent global events such as the
COVID-19 pandemic, introduces new challenges
for data governance, highlighting the need for it
to transcend boundaries – of firms, organizations
and even national governments. As noted by the
Center for Strategic and International Studies
(CSIS), in the publication “Data Governance
Principles for the Global Digital Economy”:
Dostları ilə paylaş: |