Data collection and analysis tools for food security and nutrition
participation on the part of certain segments of the
Yüklə 5,02 Kb. Pdf görüntüsü
|
participation on the part of certain segments of the population requires greater evaluation worldwide. In some instances, the digital technologies employed may be designed without obtaining user input. This is problematic when knowledge and technological divides across the stakeholders are not bridged through appropriate engagement. For instance, some farming communities in rural areas may have barriers to using technology because of the cost of adoption, lack of awareness, and accessibility or connectivity issues. In such circumstances, efforts to harness the latest digital technologies may be limited by lack of understanding of working conditions, requirements and end-user expectations. The deployment of digital technologies 58 ] DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION may sometimes take the focus and spending away from the primary objective, which is to improve the nutrition and health of the population. End users may sometimes not have the connectivity, familiarity with or digital literacy required to operate these devices. Also, continuity of efforts is needed to evaluate the benefits of technological intervention. In case of replacement or updating of a software, application or device, the advantages and benefits of the previous version and the need for improvement should be considered in tandem in order for the revised version to be useful and accepted by the end users (Johari, 2021). Thus, careful consideration of the local cultural and social context and stakeholder engagement is important for the successful adoption of newer technologies for FSN data collection and analysis. Finally, among the challenges to evaluation and decision-making in FSN, relating to SDG 2 indicators, is the lack of transparency, ownership and open access to agricultural statistics. Thus, constraints relating to ownership of and access to the information generated from data collection and analysis on the part of relevant stakeholders must be addressed. CONSTRAINTS RELATED TO THE LACK OF COORDINATION AMONG AGENCIES The collection of data on FSN indicators related to the SDGs may involve multiple agencies within a country. Fragmentation of the data collection landscape within government agencies occurs in many countries as agricultural, food and nutrition data are not collected by NSOs but by different ministries. The lack of coordinated effort among these agencies at times leads to the duplication of efforts and can unnecessarily burden financially constrained projects and initiatives. Moreover, this hinders the interoperability and linkage between datasets, which is necessary to have a holistic understanding of FSN status and its drivers in a population. For instance, some of the required data may be collected by academia, involving individual researchers whose smaller surveys may not necessarily aim to reflect the nation at large, while other data may be collected by the private agencies and may be archived behind a paywall, limiting access to the data. At the global level, much of the food security, agriculture and nutrition data are collated and disseminated by FAO. Data in the domain of health and nutrition, including those relating to maternal and child nutrition indicators such as exclusive breastfeeding, 14 are collected and disseminated by the World Health Organization (WHO) and United Nations Children’s Fund (UNICEF). However, in both these instances, the raw data comes from the individual member states or regions. As such, the quality and richness of the data typically depend on the capacity of the individual nations (OECD, 2019). The lack of coordination between national and international agencies sometimes creates gaps between objectives and delivered outcomes. For instance, 50 percent of African NSOs perceived that capacity-building programmes did not involve sufficient consultation between national and international stakeholders, and over 30 percent of NSOs worldwide expressed that the programmes did not meet their needs (PARIS21, 2018b). This demonstrates lack of sufficient country ownership of statistics capacity building programmes. Furthermore, the lack of a shared vision and accepted consensus among countries on the importance of collecting data and resistance to harmonization of indicators and data collection methodology hinder international comparisons (Veillard et al., 2010). Different UN agencies propose different methods and standards. This results in an inability to integrate and collect data across related datasets and some duplication of efforts. Some of the global constraints are reinforced by the lack of coordination between the large number of stakeholders involved and a lack of clear mechanisms for reporting and the means to deliver on their respective commitments ( SEE BOX 23 ). 14 See the WHO Tracking Tool to improve maternal, infant and young child nutrition at https://extranet.who.int/nhdtargets . 3 CONSTRAINTS, BOTTLENECKS (AND SOME SOLUTIONS) FOR EFFECTIVE USE OF FSN DATA BOX 23: A CRITICAL VIEW OF FAO STATISTICAL SUPPORT TO MEMBER NATIONS The need for better coordination of efforts is further elucidated by an evaluation of FAO statistical activities conducted by FAO itself in 2020. The aim of the evaluation was to provide Members with an assessment of FAO’s statistical contribution to agricultural and rural development and food and nutrition security from 2012 to 2018. The evaluation team concluded that FAO’s current internal statistical governance did not provide a solid basis for well-coordinated, coherent or satisfactory statistical work. This was attributed to weak enforcement of internal governance arrangements and the confusion over roles and responsibilities arising from a profusion of units and divisions conducting statistical activities (including at regional level), diluting their effectiveness. The need for FAO to better capitalize regional statistical expertise and regularly evaluate its programme resources allocated to statistical activities to ensure its appropriateness for the objectives of the work plan was recommended. The evaluation also identified that the limitation in statistical assistance provided to countries was further exacerbated by FAO’s dependence on extra-budgetary resources for statistical capacity-building, which creates uncertainty on the sustainability of this capacity-development work. Thus, despite some progress in terms of quality, the statistics produced and disseminated by FAO were deemed to be only partly compliant with its Statistics Quality Assurance Framework (SQAF). The evaluation team further recommended that FAO expedite its efforts to improve the quality of its data and IT infrastructure support and organize and enforce an integrated statistical quality management system to ensure compliance with current and new internationally accepted statistical standards and norms for all its activities (FAO, 2020c). In response to the evaluation, the Organization has taken several steps: a) FAO statistics and data for statistical purposes are governed by and already adhere to three overarching frameworks: (i) the Fundamental Principles of Official Statistics (though mostly geared toward national statistical agencies); (ii) the Principles Governing International Statistical Activities, which focus on international organizations and whose second edition (2014) was endorsed by the Director-General; and (iii) the International Statistical Institute (ISI) Declaration on Professional Ethics, which provides ethical guidance for all professional statisticians working both in academia and in national and international organizations. In particular, Principle 6 of both the Fundamental Principles of Official Statistics and the Principles Governing International Statistical Activities, as well as Principle 12 of the ISI Declaration on Professional Ethics, focus on data protection and confidentiality. b) Key FAO databases, which publish only aggregated statistical information, adhere to the open-data policy Creative Commons 3.0 Intergovernmental Organization (IGO) license. With the development of the FAO Statistics Data Warehouse (PC 132/5, paragraph 27), this license will apply to all corporate statistical databases available on the FAO website. FAO is currently also initiating discussions to upgrade to Creative Commons 4.0 IGO (CC-BY-4.0) to adhere to the Digital Public Goods Standard for Open Data, which stems from the UN Secretary-General’s Roadmap for Digital Cooperation. c) In 2019, FAO established a corporate platform for the dissemination of food and agriculture microdata (the Food and Agriculture Microdata [FAM] Catalogue) which applies the most advanced international standards and best practices in the treatment of personal data (personal data anonymization, use of statistical disclosure procedures and terms of use of microdata). d) FAO has developed corporate standards requesting the informed consent of the respondents for all surveys directly carried out by the Organization. [ 59 DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION CONSTRAINTS THAT CREATE A LACK OF TRANSPARENCY AND OF APPROPRIATE REGULATORY FRAMEWORKS A third issue in terms of institutional arrangements for effective data collection and use is the need for governments to disclose data so that it can be easily accessed and used. In some cases, a lack of political will and hesitancy to share sensitive information may prevent the collection of data and publication of results on issues such as moderate levels of food insecurity, due to the fear that they may imply challenges far greater than those perceived and accepted by the national governments (Asian Development Bank, 2013; Banik, 2016; Thow et al., 2018; Wan and Zhou, 2017). In other instances, access to food safety data may be regarded as sensitive information as this information could affect export opportunities. Another important issue in terms of data collection and dissemination is the need for strong legal and regulatory frameworks that protect human rights and privacy. This is particularly so with the increasing involvement in FSN data generation and analysis on the part of private agencies. There have been advances in methods used to collect and instantaneously process food production including agricultural data are through the use advanced sensor technologies and digital agriculture . Aquatic food is a vital source of food for people, and fish production requires constant monitoring and ready to use data. Such data access will prevent overexploitation or depletion of fish stocks and provide valuable information for effective fisheries management (Grilli, Curtis and Hynes, 2021). Moreover, smart livestock farming also uses several technologies that analyse data to improve production with reduced environmental impacts. For example, new data analytic architectures that generate farm and field level data allows farmers and stakeholders to monitor processes and make a decision for the precision livestock farming. (Fote et al., 2020). The use of these advanced technologies provides a level of granularity and immediate access to data that was lacking in traditional surveys. 60 ] 3 CONSTRAINTS, BOTTLENECKS (AND SOME SOLUTIONS) FOR EFFECTIVE USE OF FSN DATA BOX 24: SATIDA COLLECT SATIDA COLLECT is an Android application that allows for rapid and simple collection of data related to malnutrition and access to resources to support humanitarian aid organizations involved in drought and food security management. SATIDA COLLECT is a freely available, flexible and efficient mobile application that was developed using and open-source toolkit for data collection “Open Data Kit (ODK) aggregate”. SATIDA COLLECT also standardises data collection on malnutrition, socio-economic factors, access to resources, food prices, coping capacities and other related data. All assessments using SATIDA include GPS coordinates and are automatically uploaded to a database for storage. Its application programming interface (API enables data to be immediately displayed on web viewer The SATIDA database provides immediate access to the data and allows further analysis through features that enable sharing and export of assessments. In addition, it facilitates the visualization of drought risk with satellite-derived data. More importantly, from the user standpoint, it is an easy-to-use tool. SATIDA Collect was used in Central African Republic for monitoring food Security and analyse the drought risk and impacts. Note: For more information, see: https://m.apkpure.com/satida-collect/com.satida.collect.android . Source: Enenkel et al., 2015 BOX 25: TACKLING CONSTRAINTS IN FOOD COMPOSITION DATA AVAILABILITY AND QUALITY Food composition data are often used for assessment and planning of human energy and nutrient intakes, providing information for which many public health and nutrition policies and programs are based. The International Network of Food Data Systems (INFOODS) ( https://www.fao.org/infoods/infoods/en/https://www.fao.org/infoods/ infoods/en/ ) was established in 1984, aiming at stimulating and coordinating efforts to improve the quality and availability of food composition data globally. The network provides guidelines (e.g., quality assessment of data from journal articles for use in food composition tables, food matching, conversion of units), and standards (e.g., food nomenclature, terminology, classification systems, tag names), overview of food composition data management systems and software tools for dietary assessment. In addition, a comprehensive e-learning course on food composition data is available on their webpage. To circumvent the lack of availability of nutrient content of aquatic foods, that are important in diets and nutrition in many regions of the world, the INFOODS’ Global Food Composition Database for Fish and Shellfish (uFiSh) is a global database made available in Excel. uFish provides nutrient values for selected fish, crustaceans, and molluscs in raw, cooked, and processed form, covering data on proteins, minerals, vitamins, amino acids and fatty acids, primarily major finfish species. To further address the contribution from a diverse range of aquatic foods a new collaboration was launched in 2022 with multiple partners including the FAO, the University of Lancaster, WorldFish, and the Institute of Marine Resources, Norway, to increase accessibility and use of high-quality food composition data on aquatic foods to better inform public health and nutrition policies and programs based on updated and recent evidence. [ 61 62 ] Chapter 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA Colombia, Precision agriculture. © Herney Gómez [ 63 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA O ne of the most impressive and rapid developments of the last few decades has been the “data revolution” (Kitchin, 2014a) – a series of innovations that affect the way in which data are produced, managed, analysed, stored and utilized, which is dramatically changing the very nature of data and information. As eloquently put by Kitchin (2014a), in the past, data was so “time-consuming and costly to generate, analyse and interpret” that “good- quality data were a valuable commodity, either jealously guarded or expensively traded.” Nowadays “the production of data is increasingly becoming a deluge; a wide, deep torrent of timely, varied, resolute and relational data that are relatively low in cost and, outside of business, increasingly open and accessible” (Kitchin, 2014a, p.1). Navigating this torrent presents challenges and opportunities, but it is unavoidable, including for agriculture, food security and nutrition. In order to address FSN needs and opportunities associated with data, specific tasks need to be undertaken, primarily associated with the FSN data cycle (as introduced in Chapter 1), the six FSN dimensions (HLPE, 2020; Clapp et al., 2021) and some of the constraints mentioned in chapters 2 and 3. This chapter begins by identifying and defining key new and emerging digital technologies that are relevant to food systems and FSN. Next, the specific tasks associated with the FSN data cycle, FSN dimensions and data constraints are described in detail, including how specific technologies can be utilized in those tasks. The chapter closes in by highlighting risks associated with digital technologies that affect the extent to which the technologies can be successfully implemented and utilised and suggests appropriate mitigation measures. LANDSCAPE AND RELEVANCE OF NEW AND EMERGING DIGITAL TECHNOLOGIES TO FSN New and emerging digital technologies, such as big data, artificial intelligence (AI), sensors and the internet of things (IoT) and blockchain technology, feature prominently in precision agriculture, smart farming, and Agriculture 4.0 (the latter being defined as agriculture that integrates a series of technological innovations in order to enhance the agriculture value chain [Santos Valle and Kienzle, 2020]). Agriculture 4.0 has also been extended to Agri-food 4.0 – to include food supply chains (Lezoche et al., 2020). Consequently, much data are being produced, collected, processed, analysed and disseminated in the context of FSN and are influencing the FSN supply chain (Wolfert et al., 2017). Box 26 presents definitions of the key new and emerging digital technologies that are, or have a potential to be, applied to FSN. DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION BOX 26: DEFINITIONS OF NEW AND EMERGING DIGITAL TECHNOLOGIES Artificial Intelligence: Artificial intelligence (AI) is the theory and development of computer systems able to perform tasks commonly associated with human intelligence. AI includes specific fields such as machine learning, perception, robotics and natural language processing. Computer vision and deep learning can be used to support visual perception. Big data and cloud computing: Big data refers to high-volume, high-velocity, high-variety and high-veracity information assets that demand cost-effective, innovative forms of information processing for enhanced insight, decision-making and process automation. Cloud computing centralizes resources and services remotely and facilitates their use by multiple users without the need for the users to store the resources or install the services on their individual hard drives. Blockchain technology: Blockchain technology (or distributed ledger technology) refers to a decentralized, distributed record such that the data units are broken up into shared blocks that are chained together with unique identifiers. The use of blockchain technology has increased, especially due to its application in cryptocurrencies, non-fungible tokens (NFTs), smart contracts, etc. A cryptocurrency is a virtual or digital currency secured by cryptography, designed to work as a medium of exchange through a computer network that is not reliant on any central issuing or regulating authority. A non-fungible token (NFT) is a non-interchangeable unit of data stored in the form of a digital ledger that can be sold and traded. Smart contracts are contracts or agreements that can be automatically executed, enforced, controlled and documented, partially or fully, without human interaction. Crowdsensing (or community sensing) is a paradigm in which a community leverages devices with sensing and computing capabilities to collectively share data and extract information to measure and map phenomena of common interest (Kraft et al., 2020). Crowdsensing differs from the paradigm of personal sensing as, in the latter, the phenomena that are monitored belong to an individual user, while crowdsensing applies to scenarios where the phenomena of interest cannot be easily measured by a single user or device (Ganti, Ye and Lei, 2011). Crowdsourcing: Crowdsourcing is the practice of engaging a group of people (a crowd), usually via social media and the internet, to assist in collecting information, ideas, opinions or other input for a common goal, such as problem solving, innovation, etc. Decision-support system (DSS): This refers to a software-based system that gathers and analyses data from a variety of sources in order to facilitate the decision-making process for management, operations, planning or optimal solution path recommendation. Digital twin: A digital twin is a virtual representation that serves as the real-time digital counterpart of a physical object or system and that helps in decision-making. Geographic information system (GIS): GIS is a system using software tools to capture, store, analyse and visualize location-relevant data often used to study and monitor land area usage, impact of weather events, etc. Information visualization: This is the process of transforming data into an interactive, visual form that enables or triggers users to use their mental and visual capabilities to further understand and gain insight into that data. Interactive voice response (IVR): This is a technology that allows humans to interact with a computer-operated phone system using voice and dual-tone multi-frequency (DTMF) user interface, allowing them to provide and access information. Online social media: This refers to user-generated information, opinions, video, audio and multimedia that are shared and discussed over digital networks. 64 ] [ 65 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA Semantic web: Semantic web technologies enable the creation of web-based data stores, the construction of vocabularies and ontologies, and the writing of rules to process the data. At the top of the Semantic web stack is inference, which is reasoning about data using rules. Sensors and internet of things (IoT): A sensor is a device that measures a physical or chemical feature. Sensors include, but are not limited to: standard sensors (to detect soil moisture or for tracking animals, for instance), weather stations and remote sensing (for example, using satellite technology). Sensors that capture digital images or video are increasingly used to capture reality. These sensors can be fixed or mobile (mounted on tractors, robots, drones, etc.). The development of nano-computers and microcontrollers has facilitated and popularized the use of these sensors, making them accessible to a wide population. Sensors are commonly used in IoT applications. IoT refers to the network of physical objects that have sensors, software and other technologies to connect and exchange data with other devices and systems over the internet. IoT is often used together with other technologies such as machine learning, analytics, computer vision and robotics. Ubiquitous computing: Ubiquitous computing is a concept where computing is made to appear or occur anytime and everywhere. Ubiquitous computing has become widespread, especially through mobile computing, where end users carry their devices (such as mobile phones) and use them in everyday activities and contexts. Mobile computing applications can be based on Short Message Service (SMS), Unstructured Supplementary Service Data (USSD), chatbots, computer-assisted telephone interviewing (CATI), and other forms of applications, such as Open Data Kit (ODK)-based technologies. Virtual reality and augmented reality: Virtual reality (VR) is a computer-generated simulated environment with objects and scenes that seem real, making the user feel immersed in the simulated environment. Augmented reality (AR) is an interactive experience of a real-world environment where the objects in the real world are enhanced by computer-generated information and features. New and emerging digital technologies can support all the stages of the data cycle for FSN decision-making. They can also support the FSN dimensions and address some of the data- related constraints mentioned in chapters 2 and 3. The following sections describe specific tasks associated with each stage of the FSN data cycle and relevant new and emerging digital technologies for each task, with examples. FSN dimensions and constraints (referred to in the introduction and in Chapter 3) associated with the specific tasks are also mentioned. DEFINE/REFINE EVIDENCE PRIORITIES AND QUESTIONS Among the tasks associated with this stage of the data cycle is assessing options and proposing priorities and questions . As explained by Yoshida (2016), for example, various methods are used to set priorities in health and nutrition research. For instance, networks such as the Child Health Nutrition Research Initiative (CHNRI) and the James Lind Alliance (NIHR - National Institute for Health Research, 2021) gather information from experts (for example through the Delphi technique and through focus group discussions) and consolidate the expert opinions in order to set priorities. This approach, based on expert opinions, can be supported using digital technologies such as: Short Message Service (SMS), Unstructured Supplementary Service Data (USSD), chatbots, crowdsourcing, machine learning, Open Data Kit (ODK)-based technologies, Interactive Voice Response (IVR) and other mobile applications. Such technologies may also help FSN actors contribute to and articulate priorities and weigh the different options (using machine learning, for example) thereby potentially improving clarity on priorities. Wazny et al. (2019), for instance, used the CHNRI method to set research priorities for maternal and child health and nutrition in India, using crowdsourcing to collect research ideas from a network of 66 ] DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION child health experts from across India. These approaches and technology may help to address the constraint pertaining to lack of clarity on how to prioritize (mentioned in the introduction). It is important to note that while the approach used by the CHNRI includes the collection of many research ideas from different sources (researchers, policymakers and programme managers), this process is executed in order to define/refine evidence priorities and questions and is therefore part of the first stage of the data cycle, rather than being part of the next stage in the data cycle of reviewing, consolidating, collecting and curating data . REVIEW, CONSOLIDATE, COLLECT AND CURATE DATA The data cycle stage concerned with reviewing, consolidating, collecting and curating data includes a number of specific tasks that can be supported by new and emerging digital technologies. One of these tasks is supporting the collection and production of FSN-relevant data, a task that can take many forms, including collecting FSN data from respondents and complementing self-reported data. Digital technologies that can support the collection of FSN data from respondents include crowdsourcing , crowdsensing , online social media , SMS, USSD, chatbots, ODK-based technologies, IVR and other forms of mobile applications. Information collected from respondents using these technologies can include a wide variety of data relevant to any of the six FSN dimensions. Respondents can report about incomes, expenditures, prices and the status of physical transport and communication infrastructure – information is relevant to the FSN dimension of access . For example, Ochieng (2019) describes a pilot study conducted in Malawi to crowdsource farm gate prices for pigeon peas and chickpeas through the Farm Radio Trust platform. Farmers reported the prices and locations at which they had sold their produce. Respondents can also report about feeding practices, food preparation, food safety, dietary diversity and health-seeking behaviour. Such information is relevant to the FSN dimensions of utilization . For example, De Choudhury, Sharma and Kiciman (2016) conducted a study to estimate the quality of available foods in different geographical locations using data from 3 million food-related posts shared on social media. The study found that the foods in social media posts shared by people located in food deserts were higher in fat, cholesterol and sugar intake and lower in protein and fibre. Another effort by Shah et al. (2020), used natural language processing and machine learning algorithms to collect and analyse data from Twitter in order to assess Canadian’s health and nutritional habits. The model classified food and non-food posts and provided information (such as caloric intake vs energy expenditure) of Twitter posts per province as well as foods and activities most tweeted about per province. Respondents can report activities and events relevant to the FSN dimension of sustainability , such as those that are related to the environment and climate. For example, MIT’s Climate CoLab ( https://www.climatecolab.org ) is an effort that taps into the collective intelligence of people from all around the globe to address societal problems, starting with climate change. Climate CoLab provides an open problem-solving platform through which thousands of people work on and assess plans to reach global climate-change goals. Respondents (including farmers, veterinary officers and agricultural extension officers) can also report and help monitor the presence of pests and diseases which damage sources of food. For example, the Agritask agronomic platform, developed by the International Center of Insect Physiology and Ecology (ICIPE) and Tel Aviv University (TAU), provides a mobile application for field scouts and lead farmers to report pests from the field. The operations of the platform span four counties in Kenya, covering approximately 20 000 small subsistence farms. 15 15 See https://start.agritask.com/wp-content/uploads/2020/10/ Agritask-ICIPE-Case-Study-Final.pdf . [ 67 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA The task of collecting and producing FSN-relevant data may also involve complementing self- reporting data. Machine learning can be used for this purpose. For example, Schmidhuber et al. (2018) have used machine learning algorithms to extract dietary intake data from the Global Burden of Disease study and developed predictive models that estimate the consumption of each nutrient based on its national availability. Such data can inform initiatives to address nutritional needs of specific populations in the context of particular food systems. Moreover, the task of collecting and producing FSN-relevant data may involve the automated collection of data pertaining to FSN entities such as: agricultural fields, weeds, pests, diseases, natural phenomena (such as weather) and natural food resources (e.g., wild foods, including fish). Digital technologies that can support the automated collection of data pertaining to FSN entities without involving respondents directly include: remote sensing technologies, GIS, robotics, IoT and digital twins . For example, WFP DataViz is a data visualization platform that provides interactive geographical and graphical information through Hunger Hub, Seasonal Explorer, Economic Explorer, Interactive Reports and Thematic Dashboards. Remote sensing data comes from the MODIS (moderate-resolution imaging spectroradiometer) instrument on board the NASA satellites Terra and Aqua. The remote sensing data are updated on a regular basis. The raster data (pixelated data where each pixel corresponds to a particular geographic location) are processed, aggregated and geo-referenced in order to present an easy-to-understand visualization ( https://dataviz.vam.wfp.org/ ). Another example is Flybird Innovations, a social impact agricultural enterprise in India. Flybird has developed Siri, is a smart irrigation controller. Siri manages water and fertilizer application to crops and plants. Siri has sensors that collect data on soil moisture, temperature and humidity in order to prevent under- and over-irrigation and fertilization. Flybird also collects basic demographic information on its farmers, as well as geographic and crop data. This information enables Flybird to predict water requirements and optimal fertilization for farmers’ crops ( http:// www.flybirdinnovations.com/ ). As the examples described in this section demonstrate, many digital technologies can facilitate the task of supporting the collection and production of FSN-relevant data . This, in turn, can help address the constraint pertaining to lack of available data , which was mentioned in the introduction. Moreover, new and emerging digital technologies can enhance the collection, storage and processing of qualitative data in the form of images, videos, audio recordings and text. Online social media, crowdsourcing and other mobile computing-based applications, for instance, can enable the collection of qualitative data. Big data and cloud computing can support the storage of qualitative data. Machine learning, through sentiment analysis for instance, enables the analysis of qualitative data. Online social media and information visualization enable the dissemination of qualitative data (Kanter and Gittelsohn, 2020). This can contribute to addressing the constraint pertaining to over- reliance on quantitative data , mentioned in the introduction. Another task associated with the data cycle stage of reviewing, consolidating, collecting and curating data is linking, integrating, aggregating and enriching data from different sources . Digital technologies that can support this task include: semantic web , big data and digital twins. Examples of efforts in this regard include: FoodOn, CGIAR’s Crop Ontology, FAO’s AGROVOC, Wageningen University & Research’s Digital Twin projects, BeeZon‘s Virtual Bee Consultant, and the CGIAR Platform for Big Data in Agriculture, which are described in Box 28. These digital technologies (in particular semantic web and big data) can contribute to improving access to data – mentioned in Chapter 2 as a key constraint. Semantic web can in fact support harmonization and interoperability of data and systems. Moreover, interoperability could also facilitate efforts around open- source tools and materials, which in turn can further contribute to data access. Open-access initiatives, open-source efforts and digital DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION technologies (such as big data, crowdsourcing and mobile computing) can also help to alleviate constraints associated with inadequate infrastructure and insufficient resources and capabilities (mentioned in Chapter 3) which arise in other stages of the data cycle. BOX 27: EXAMPLES OF EFFORTS THAT SUPPORT DATA CONSOLIDATION FoodOn is an ontology that describes common foods from around the world. The ontology can be used to construct statements about food, which can be entered in a database and queried or reasoned about. FoodOn focuses on foods for humans and domesticated animals. It contains animal and plant food sources, food categories and products, and other facets such as preservation processes, contact surfaces and packaging (Dooley et al., 2018). CGIAR’s Crop Ontology provides descriptions of agronomic, morphological, physiological, quality and stress traits along with a standard nomenclature for composing the variables. The ontology enables digital capture, aggregation and integration of crop trait data, as well as comparisons across farmers, breeders, scientists and other communities, through surveys with citizen science tools. As of 10 November 2020, the CGIAR website reported that the ontology comprised 4 235 traits and 6 151 variables for 31 plant species (www.cropontology.org) and supported the generation of FAIR (findable, accessible, interoperable and reusable) data ( https://bigdata.cgiar.org/digital-intervention/crop-ontology-2/ ). FAO AGROVOC ( https://www.fao.org/agrovoc/about ) is a multilingual and controlled vocabulary designed to cover FSN-relevant concepts, terms, definitions and relationships. The concepts are used to support unambiguous identification of resources and standardization of indexing processes, and to make searching more efficient. Each concept also has terms used to express it in various languages. AGROVOC consists of over 39 800 concepts and over 929 000 terms in up to 41 languages. Wageningen University & Research’s Digital Twin projects , which are still under development, comprise: Virtual tomato crops; Me, my diet and I; and Digital Future Farm ( https://www.wur.nl/en/newsarticle/WUR-is-working-on- Digital-Twins-for-tomatoes-food-and-farming.htm ). he virtual tomato crops project is developing a digital twin of a real tomato crop in a greenhouse – a 3D simulation model that is fed in real-time with sensor information from a real greenhouse. The interactions between the specific characteristics of the tomato crop, the environmental factors and crop management measures are all simulated in the virtual crop. Since the model is linked to a real tomato crop in a greenhouse, it is possible to continually refine predictions and thus make better choices for the real crop. It is anticipated that once the model is completed, growers can use it as a decision-support tool for growing real tomato crops in a greenhouse. For example, it will allow growers to predict the effect of a crop management measure on crop harvest and financial yield and thus make a decision for the real tomato crop based on that prediction. Virtual Bee Consultant by BeeZon ( www.beezon.gr ) is a digital twin solution of bee colonies involving a real-time continuous apiary monitoring system that enables beekeepers to remotely monitor their apiaries and make smart management decisions with minimal in-person interaction. The solution is based on a GPS-based tracking system and real-time data from various sensors (measuring humidity, exterior and interior temperature, brood temperature and weight). Specifically, beekeepers can remotely monitor and act upon the following aspects: timing of nectar flows; identifying the presence of diseases, pest infection, pesticide exposure and toxicity; insight into colony status, dynamics and hygiene; identification of queenless and swarming states; management of food storage reserves; antitheft mechanisms and tracking systems; and notification systems tailor-made by the user (Verdouw and Kruize, 2017). CGIAR’s Platform for Big Data in Agriculture ( https://bigdata.cgiar.org/ ) aggregates data from various different sources. This is facilitated through the Global Agricultural Research Data and Innovation Network (GARDIAN, https://gardian.bigdata.cgiar.org/ ), which enables searches across all CGIAR repositories and connection to more datasets from strategic partners. 68 ] [ 69 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA Enabling respondents to assist in cleaning up data is another task associated with the data cycle stage of reviewing, consolidating, collecting and curating data. Digital technologies that can enable respondents to assist in data clean-up include crowdsourcing, crowdsensing, online social media, other forms of mobile applications and IVR. For instance: crowdsourcing efforts similar to the ones presented by Chu et al. (2015) could be applied in FSN. Furthermore, data validation, verification, authentication, traceability and transparency is another task associated with the data cycle stage of reviewing, consolidating, collecting and curating data. Digital technologies can support this task. For instance, ODK-based technologies support validation of user input captured through online forms and other types of user interfaces. Moreover, digital technologies, such as blockchain technology , machine learning, crowdsourcing, crowdsensing, online social media, mobile computing and IVR, are increasingly supporting validation, verification, authentication, traceability and transparency through more sophisticated means. Specific examples of these technologies include Barilla’s blockchain system, the Blockchain Supply Chain Traceability Project, WFP Building Blocks, and AgUnity’s blockchain application, as described in Box 28. Still on blockchain technology, cryptocurrencies are being experimented for adoption in FSN. One example is by AgriDigital, a technology provider for the grains industry that connects physical inventory, supply chain data and finance ( www.agridigital.io/ ). In December 2016, AgriDigital executed the world’s first sale of 23.46 tonnes of grain between farmer and buyer via blockchain. AgriDigital has so far transacted more than 1.6 million metric tonnes of grain (Sylvester, 2019). Furthermore, the Colombian delivery app Rappi, which offers on-demand deliveries of food and other goods across Latin America, launched a cryptocurrency payment pilot programme in Mexico in April 2022 (Reuters, 2022). Moreover, Burger King has been piloting transactions in cryptocurrencies in Germany, the Netherlands and Venezuela, McDonald’s has been experimenting with this in El Salvador, and KFC has been piloting this in Canada (Traders of Crypto, n.d.). DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION BOX 28: EXAMPLES OF THE APPLICATION OF BLOCKCHAIN TECHNOLOGY TO FSN DATA Barilla’s blockchain system: Barilla collaborated with IBM in 2018 to develop a blockchain system to add transparency and traceability to its pesto production cycle ( https://cryptonews.net/en/editorial/technology/icons-of- italian-business-opt-for-blockchain/ ). Through the blockchain system, customers can verify the details of a product, including cultivation, treatment, harvesting, transportation, storage and quality control (Sylvester, 2019). Digital technologies are therefore relevant to the FSN dimension of utilization. Barilla’s blockchain system also shows the possibility of using digital technologies to authenticate and promote transparency of FSN data. For instance, to authenticate and promote transparency of measures, indicators and scales. The Blockchain Supply Chain Traceability Project: This project was initiated in 2018 by World Wildlife Fund (WWF) New Zealand, WWF Australia, WWF Fiji, ConsenSys, TraSeable and Sea Quest Fiji Ltd. The project uses blockchain technology to track tuna towards stamping out illegal fishing and human rights abuses in the tuna industry. Through blockchain technology, a simple scan (for instance through a smartphone using a QR code) of tuna packaging tells the story of a tuna fish, including where and when the fish was caught, by which vessel and the fishing method used) ( https://www.wwf.org.nz/what_we_do/marine/blockchain_tuna_project ). Blockchain technology can therefore support the measurement of sustainability in FSN. WFP Building Blocks: This is a blockchain solution for authenticating and registering transactions. The blockchain solution was tested as a proof-of-concept by WFP in January 2017 in Sindh Province, Pakistan. Four months later, WFP launched a pilot project covering 10 000 Syrian refugees in Azraq refugee camp. In January 2018, the pilot was extended to cover 100 000 refugees living in camps (Sylvester, 2019). The solution enables people to receive different types of assistance from multiple humanitarian organizations at once, thus reducing the complexity of accessing humanitarian support. At the same time, no sensitive information is stored anywhere on Building Blocks. Since 2017, the solution has been scaled to provide USD 325 million worth of cash transfers to 1 million refugees in Bangladesh and Jordan. It is considered the world’s largest implementation of blockchain technology for humanitarian assistance ( https://innovation.wfp.org/project/building-blocks ). WFP Building Blocks demonstrates that blockchain technology can support the FSN dimension of access. AgUnity’s blockchain application: AgUnity has developed a smartphone application to tackle the financial and digital exclusion of remote smallholder farmers and rural communities using blockchain technology. The smartphone application helps farmers plan, sell produce, buy inputs and track everyday transactions. In a project funded by USAID, AgUnity has partnered with Virginia Tech ( in the United States of America) and Egerton University (in Kenya) and customized the smartphone application to increase the flow of African indigenous vegetables to end consumers to help increase food and nutrition security in the western part of Kenya ( https://www.einnews.com/pr_ news/541948521/exploring-the-use-of-blockchain-technology-to-improve-food-security-in-western-kenyahttps:// www.einnews.com/pr_news/541948521/exploring-the-use-of-blockchain-technology-to-improve-food-security- in-western-kenya ). As described in this subsection, digital technologies can assist in data validation, verification, authentication, traceability and transparency and in enabling respondents to assist in cleaning up data . Thus, digital technologies can contribute to data quality . ANALYSE DATA USING APPROPRIATE TOOLS One of the key tasks associated with the data cycle stage of analyzing data is analyzing, detecting and predicting FSN-relevant aspects and 70 ] [ 71 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA entities , such as: food production; food supplies, food aid and food stock levels; markets; prices; dynamics of net trade; inequity factors; occurrence of adverse conditions; feeding practices; trade; pests; diseases and nutrition aspects. The task of analyzing, detecting and predicting FSN- relevant aspects and entities is relevant to any FSN dimensions depending on the FSN aspects under consideration. For instance, if the FSN aspects under consideration pertain to food production, then the task of analyzing, detecting and predicting FSN-relevant aspects and entities will be relevant to the FSN dimension of availability . Machine learning, big data and analytics can greatly support the task of analysing, detecting and predicting FSN-relevant aspects and entities . For example, Talukder and Ahammed (2020) used machine-learning algorithms, specifically Random Forest (RF), Support Vector Machines (SVM) and K-Nearest Neighbor (KNN) algorithms, to process dietary patterns or nutrient and food intake data in order to predict malnutrition among children under five years of age in Bangladesh. Another machine learning algorithm, namely linear regression, was used to identify risk factors for stunting, underweight and wasting among under-five children in Bangladesh. In a similar effort (Rahman et al., 2021), RF exhibited high accuracy in predicting malnutrition among under-five children in Bangladesh. Another effort implemented by Kwon et al. (2020) uses a machine-learning algorithm to identify risk factors for low muscle mass based on nutritional and health-related factors among men and women. The algorithm generated five separate clusters for men and women based on age, total energy, carbohydrate ratio, protein ratio, fat ratio, smoking habits, alcohol consumption, physical activity and number of chronic diseases, yielding similar characteristics among each cluster. Another machine learning algorithm, namely logistic regression, was subsequently used to analyse the associations between each of the nine variables and low muscle-mass index, hence identifying risk factors within each cluster. A similar effort, described by Zeevi et al. (2015), uses a machine-learning algorithm integrating blood parameters, dietary habits, anthropometrics, physical activity and gut microbiota to predict personalized postprandial glycaemic response to real-life meals. Yet another effort that supports the task of analysing, detecting and predicting FSN- relevant aspects and entities is PlantVillage Nuru, 16 an on-farm pest and disease identification system. Deployable as a mobile application, PlantVillage Nuru can help smallholders detect, identify and manage cassava diseases. The system development team annotated more than 200 000 cassava plant images, identifying and classifying diseases to train a machine-learning model. As of June 2020, the mobile application had been downloaded and used in more than 40 countries and had generated more than 18 000 reports from users. Another task related to the data cycle stage of analysing data is mapping and monitoring FSN aspects and entities , such as agricultural fields, infrastructure, livestock herds, natural phenomena and natural food resources (including wild foods and fisheries resources). The collection and linking of the underlying data could be related to what was described earlier regarding review, consolidation and curation of data. Digital technologies that can support mapping and monitoring of FSN aspects and entities include: AI, information visualization, IoT, GIS, satellite technologies and digital twins. For example, PeskAAS, which is an open- source monitoring and analytics application that enables the collation, classification, analysis and visualization of data pertaining to small- scale fisheries catch and effort. Through the application, fishers themselves, managers and researchers can gain insights into a fisher’s experience of fishing efforts, fisheries status, catch rates, economic efficiency and geographic preferences and limits that can potentially guide management and livelihood investments. The application primarily uses classification, analytics and an information visualization dashboard that was codesigned with fisheries 16 See https://bigdata.cgiar.org/divi_overlay/plantvillage-nuru/ . 72 ] DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION experts and government managers (Tilley, Lopes and Wilkinson, 2020). PeskAAS is relevant to the FSN dimension of availability, among others. Another example is the GEOGLAM Crop Monitor for the G20 Agricultural Market Information System (AMIS) ( SEE BOX 2 ). AMIS provides open, timely, science-driven information on crop growing conditions, status and agroclimatic factors likely to impact global production. It focuses on the major producing and trading countries for the following four primary crops: wheat, maize, rice and soybean. The information is presented as reports that include interactive visualizations (Becker-Reshef et al., 2019). AMIS is relevant to the FSN dimension of stability. Another example which supports the task of mapping and monitoring FSN aspects and entities and which is relevant to the FSN dimension of stability is the Integrated Food Security Phase Classification (IPC) Mapping Tool. The tool uses interactive and customizable maps to visualize data. Each country is colour-coded by its latest IPC classification for both acute food insecurity (AFI) and chronic food insecurity (CFI) scales ( https://www.ipcinfo.org/ipc-country- analysis/ipc-mapping-tool/ ). The East Africa Drought Watch is another example relevant to the task of mapping and monitoring FSN aspects and entities. It is a near- real-time platform that uses earth observation and weather data to monitor drought in East Africa. The platform has been adapted from the European Drought Observatory and customized to the East Africa region. The East Africa Drought Watch is part of the Intra-ACP (African, Caribbean, and Pacific) Climate Services Project. The platform has involved collaboration with the Drought group of the Natural Disaster Risk Unit at the Joint Research Centre of the European Commission. The platform monitors several indicators, including Standardized Precipitation Index (SPI), Soil Moisture Anomaly (SMA) and anomalies of satellite-measured FAPAR (Fraction of Absorbed Photosynthetically Active Radiation) (See https://droughtwatch.icpac.net/ https://droughtwatch.icpac.net/ ). Another example which supports the task of mapping and monitoring FSN aspects and entities is the online platform Global Forest Watch (GFW). The platform provides data and tools for monitoring forests through which users can access near-real-time information about where and how forests are changing around the world using dashboards of maps and visualizations ( https://www.globalforestwatch. org ). Since forests support an ecosystem that can sustain food production in the long term, through climate change mitigation, soil formation, soil erosion control and biodiversity conservation (Meybeck et al., 2021), the GFW online platform is relevant to the analysis of sustainability of FSN. TRANSLATE DATA INTO RESULTS, INSIGHTS AND CONCLUSIONS Digital technologies can be used to support various tasks associated with translating data into results, insights and conclusions on FSN. These include aiding the presentation of data to users by rendering it easy to understand . Information visualization is one of the primary technologies for supporting the understandability of data in any of the six FSN dimensions. A notable example is the Food Systems Dashboard, presented in Chapter 2, which combines data from different sources to facilitate understanding, comparison and decisions on food systems ( https:// foodsystemsdashboard.org ). Another example that uses information visualization is the ICES Marine Food Stock Assessment Database, whose marine food stock assessment information is presented using graphs and tables ( https://www.ices.dk/ data/assessment-tools/Pages/stock-assessment- graphs.aspx ). DISSEMINATE, SHARE, REVIEW, DISCUSS RESULTS, REFINE INSIGHTS AND CONCLUSIONS Digital technologies can be used to support a range of specific tasks linked to the stage of the data cycle concerned with dissemination, sharing, review, discussion of results and refinement of insights and conclusions . [ 73 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA One of the tasks associated with this data cycle stage is making data accessible. New and emerging digital technologies, such as big data, machine learning, semantic web, visualization, crowdsourcing, online social media and mobile computing, can be used to make data accessible. In this way, digital technologies can contribute to addressing of the constraint associated with lack of availability and access to data. One example a relevant digital application is mKisan ( https:// mkisan.gov.in ), a mobile application that makes agroadvisory data accessible to smallholder farmers in India. The application was one of the first examples in India of a platform through which smallholders can access agrometeorological and market price information and receive related advice on crops and livestock, thus making mKisan especially relevant to the FSN dimension of agency. Other tasks associated with the data cycle stage of data dissemination, sharing, review, discussion of results and refinement of insights and conclusions are supporting efficient communication, wide distribution and discussion. Online social media can play a key role in supporting the foregoing task. For example, WhatsApp was used in 2016 by project implementers from the ministries of health and agriculture in selected counties of Kenya to share additional monitoring data in the form of photographs, videos and texts regarding farm status, sales, activities that the implementers carried out, etc. As a result, information delays were reduced due to collapsed reporting hierarchies, and project monitoring-related costs were reduced by 51 percent. A shared understanding on the part of different actors on the project’s indicators, reporting timelines and data collection guidelines improved the quality of continuous monitoring data (Chesoli, Mutiso and Wamalwa, 2020). Digital technologies can therefore improve timeliness in the dissemination of data (and also in other stages of the data cycle). Digital technologies (such as robotics, machine learning and DSS) can enhance efficiency of specific FSN activities as well as addressing some human resource constraints (such as responsible digital automation and user- and context-adaptive digital systems). Another task relevant to this data cycle stage is promoting transparency, traceability and accountability. Digital technologies that can support this task include crowdsourcing, crowdsensing, online social media, information visualization and blockchain technology. (This task is also related to the task of validation, verification, authentication, traceability and transparency which was mentioned earlier). An example relevant to this task is a feasibility study carried out by Global Pulse in partnership with FAO and WFP in which crowdsourcing was used to track food prices in near-real-time in Nusa Tenggara Barat, one of Indonesia’s poorest provinces ( https://www.unglobalpulse.org/ project/feasibility-study-crowdsourcing-high- frequency-food-price-data-in-rural-indonesia/ ). The area comprised almost exclusively informal, cash-only markets and stalls, where availability of other data sources was limited. The study involved local citizen reporters who submitted food price reports via a customized mobile phone application. One of the findings was that crowdsourcing, which captures high-frequency data on local trends, is best deployed in areas where traditional data capture methods are difficult, impractical or costly due to insecurity, food price volatility and geographic dispersion. This example shows that digital technologies can contribute to enhance livelihoods and, thus, access to food. USE RESULTS, INSIGHTS AND CONCLUSIONS TO MAKE DECISIONS One of the tasks associated with the data cycle stage of using results, insights and conclusions to make decisions is profiling food security and nutrition entities and using the resultant data to gain insights for decision-making . FSN entities that can be profiled include: equipment, animals, crops, food, relevant people (for example, subjects such as farmers and consumers), natural phenomena, etc. New and emerging digital technologies that can support this task include AI, big data, information visualization and digital twins. For example, Destination Earth, (or DestinE) (see https://digital strategy.ec.europa. eu/en/library/destination-earth ), an initiative of the European Commission, is developing a digital 74 ] DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION model of the earth (a digital twin) to monitor and predict natural phenomena and related human activities towards supporting sustainable development and tackling complex environmental challenges. The digital twin will be a digital replica of the earth system and will be built based on the domains of earth science. DestinE is expected to help policymakers simulate and monitor the earth’s system developments (land, marine, atmosphere, biosphere) and human activities; anticipate environmental disasters and resultant socioeconomic issues in order to protect lives and avoid major economic downturns; and enable the development and testing of scenarios for guiding more sustainable development. These efforts can contribute to providing data on sustainability. If profiling data are used to track relevant food security and nutrition indicators, they can contribute data on food utilization. For example, one of Wageningen University & Research’s Digital Twin projects (mentioned earlier) is the Me, my diet and I project (see https://www. wur.nl/en/newsarticle/WUR-is-working-on- Digital-Twins-for-tomatoes-food-and-farming. htm ). This project is bringing together human nutrition, health, AI and social science experts to build a personalised digital twin to predict the rise in blood sugar (glucose) and in blood fat (triglyceride) after a meal. The project is also expected to provide individualized nutritional advice based on personal data, such as body mass index, age, body fat distribution and blood pressure. Digital technologies such as big data, machine learning, semantic web, visualization, crowdsourcing, online social media and mobile computing can provide users with information and resources that can guide them in making their own decisions, thereby supporting the FSN dimension of agency. For example, FoodSwitch, which is a mobile application that provides users with easy-to-understand nutrition information and support the selection of healthier choices when shopping for food. It allows users to scan the barcodes of food and drink products and instantly see whether they are high (red), medium (amber) or low (green) in fat, saturates, sugars and salt. It also searches the database for similar but healthier alternative products, facilitating the switch to healthier food choices (Dunford et al., 2014). The application uses crowdsourcing to obtain nutritional information on additional food products. The application has so far been launched in Australia; China; China, Hong Kong SAR; Fiji; India; Kuwait; New Zealand; South Africa; the United Kingdom of Great Britain and Northern Ireland and the United States of America (see https://www. georgeinstitute.org/projects/foodswitch ). Another example is the mobile and web platform by and for Inuit, called SIKU ( https://siku.org/about ). It provides tools and services for indigenous knowledge pertaining to aspects such as weather conditions, sea-ice safety, wildlife sightings and sharing information about hunting exploits. Some of the tools supported by the platform use digital technologies such as online social media and geographical mapping of sea ice using Google Street View. Another related effort is Digital Green ( https://www.digitalgreen. org/ ), a development organization that aims to empower smallholder farmers to lift themselves out of poverty by harnessing the collective power of technology and grassroots-level partnerships using various tools, such as social media and mobile applications. It is worth noting that mobile applications can play a key role in empowering smallholder farmers and other vulnerable FSN stakeholders, for instance, through mobile financial services. According to the Global System Mobile Association, mobile financial services can be beneficial to smallholder farmers in various ways, including time and cost savings, convenience and efficient cash management. Moreover, mobile money technology can enable agribusiness companies to lower the costs of withdrawing, transporting and securing cash; facilitate real-time payments across multiple locations and mitigate risks associated with handling cash, such as theft and fraud (Arese Lucini, Okeleke and Tricarico, 2016). Furthermore, recent studies have observed that mobile money adoption can have a positive effect on farm input use, farm output and welfare of smallholder farmers (Abdul-Rahaman and Abdulai, 2022; Peprah, Oteng and Sebu, 2020). [ 75 4 NEW AND EMERGING DIGITAL TECHNOLOGIES FOR FSN DATA RISKS ASSOCIATED WITH DIGITAL TECHNOLOGIES FOR FSN AND THEIR MITIGATION As noted in Chapter 3, some of the risks and issues inherent in data collection and analysis can be even more relevant to digital technologies, while some typologies of risks are exclusive to these technologies. This section describes various risks associated with the new and emerging digital technologies for FSN. It also proposes measures that can be taken to mitigate these risks. ETHICS, DATA PROTECTION, TRUST, JUSTICE AND IDENTITY There are various ethical concerns associated with digital technologies, as they can be used to undertake tasks in a manner that undermines or overrides personal judgment. While in certain specific situations (to avert disaster, for instance) undermining or overriding autonomous individual choices may be beneficial, there are scenarios where the capacity to do this may be used maliciously. For instance, AI can be used to manipulate user behaviour in a way that undermines autonomous rational choice. Users’ intense interaction with AI systems enables the latter to collect a great deal of knowledge about the users. Notwithstanding the potential benefits of acquiring and using such knowledge, algorithms can be used to target users and, therefore, influence them (Narayanan et al., 2020). This manipulation often uses dark patterns, whereby user interface design choices coerce, steer or deceive users into making decisions that, if fully informed and capable of opting for alternatives, they might not make. For example, through AI, social media can aggressively advertise unhealthy food to vulnerable categories of users, such as children and adolescents (Freeman et al., 2014). While digital technologies can be used to support and promote human rights and justice in FSN, there are situations where inconsiderate digital automation (such as through AI and robots) may create conflict with such norms. As Yeung (2018) notes, the use of algorithmic decision-making by AI systems can contribute to discrimination and threaten human rights in various ways, for instance when there are biases inherent in algorithmic decision-making AI systems. This can happen if the developers of the algorithm are (consciously or unconsciously) biased, if biases are built into the model upon which the systems are built or are present in the training data or in the input data (European Union Agency for Fundamental Rights, 2019), or if they are introduced when such systems are implemented in real-world settings. These biases might create or reinforce existing discrimination. While acknowledging that technology-based decision-making can enhance the accuracy, effectiveness and efficiency of law enforcement, General Recommendation No. 36 on Preventing and Combating Racial Profiling (24 November 2020) by United Nations Committee on the Elimination of Racial Discrimination, also points out that big data and AI tools may reproduce and reinforce already existing biases and lead to even more discriminatory practices ( https://www.ohchr.org/en/hrbodies/cerd/ pages/cerdindex.aspx ). Another way through which the use of algorithmic decision-making AI systems may contribute to discrimination and threaten human rights, is when there is lack of transparency of the complex digital technology behind the systems (Yeung, 2018). The foregoing issue limits the ability of users to Yüklə 5,02 Kb. Dostları ilə paylaş:
|