Figures
3.1 Private stack, state stack, and public stack illustrations 29
3.2 Public, private, and state stack continuum 29
4.1 Two systems in the brain that control behaviour (Kahneman, 2011) 54
4.2 Original Technology Acceptance Model (Davis, 1986) 57
4.3 The diffusion process (Rogers, 1995) 58
4.4 The Strategies and Motives for Resistance to Persuasion (SMRP) Framework (Fransen et al., 2015) 59
4.5 Self-Determination Theory (Ryan and Deci, 2000) 62
5.1 Illustration of the Socio-Technological Feedback Loop from a human-machine interaction perspective (Aliman and Kester 2022a) 75
5.2 Illustration of the Socio-Technological Feedback Loop from a system life-cycle perspective 76
6.1 Equipped with a torch and landing net, a group of nature enthusiasts visit pools in and around the city of Eindhoven 83
6.2 The number of active environmental citizen science projects through the years, according to the inventory maintained by the European Commission (JRC, 2018) 85
6.3 Impact of all citizen science projects on SDGs 88
7.1 Technology tends to reduce our connection with nature 99
7.2 Theoretical model of the mutual influence between system level and daily life level 100
7.3 Case study 1, walking apps may be designed to stimulate people to connect more strongly to nature, e.g. by providing information and stories 108
7.4 Case study 2, citizen science organized by the Dutch Butterfly Conservation foundation. Buckets with LED-light are used by citizen scientists to record moths in the Netherlands, while butterflies may be recorded along transects 111
7.5 Case study 3, IJsselstein is a municipality in the Netherlands with a large collection of fruit trees in the public green space. Technology, including a website and GIS application, provides insight into the uniqueness of this collection and helps to coordinate activities associated with it 113
8.1 The value profiles of two people that participated in the moral food lab. Note that the size of the words is in no way correlated with the number of times the words were mentioned 129
8.2 The average percentage of values people spoke in, grouped by how they responded to question B 130
10.1 Typical AI lifecycle model 155
10.2 The computational effort of AI models increases according to Moore’s Law of the AI era 157
10.3 Carbon intensity for an assortment of locations 159
10.4 Benchmark of CNN models for image classification 160
10.5 Clean data improves prediction accuracy 161
10.6 Clustering the E-waste dataset into device groups 164
10.7 Training process of a neural network 166
11.1 The five dimensions of sustainable software engineering 171
11.2 ISO 25000 quality model 173
11.3 Energy monitoring setup for mobile app development 175
11.4 Development process for AI-enabled systems, including a data and model (ML) loop 182
11.5 Green AI at the root of Trustworthy AI 182
12.1 Leaflet Media Gym, Studio Cream on Chrome 187
12.2 Screenshot of taste workshop by E-missions 191
13.1 Examples of plants in an urban environment. a) Bird’s-foot Trefoil (Lotus corniculatus), b) Ivy-leaved Toadflax (Cymbalaria muralis), c) Kidney Vetch (Anthyllis vulneraria), d) White Clover (Trifolium repens), e) Dandelion (Taraxacam officinale), f) Yarrow (Achillea millefolium), g) Common Poppy (Papaver rhoeas), h) Ground Elder (Aegopodium podagraria), i) Wallflower (Erysimum cheiri) 198
13.2 Impact of green design of private urban gardens on quality of living environment. Left: tiled backyard with overheated owner with irritated respiratory tract due to allergenic pollen released by ornamental olive shrubs. Right: backyard filled with non-allergenic trees, shrubs and herbs, providing shade, water retention, food to wild animals and a general feeling of well-being to owners 201
13.3 Example of urban trees encouraging bird safaris. Migrating Bohemian Waxwings (Bombycilla garrulus) foraging for berries in a tree planted along the canal of a typical Dutch historical urban center with bird watchers enjoying the scene, while keeping a respectful distance 202
14.1 Challenges of flower identification ‘in the wild’: 1) viewpoint variations (Papaver rhoeas); 2) occlusion (Ranunculus repens); 3) clutter (Achillea millefolium); 4) light variation (Leucanthemum vulgare); 5) deformations (Bellis perennis); 6) intra-class variation (Ficaria verna); 7) inter-class similarity (Bellis perennis, Leucanthemum vulgare, Matricaria chamomilla) 209
14.2 Image recognition and object detection comparison. Image recognition (first and third) labels the entire image, while object detection (second and forth) localizes objects in an image by drawing bounding boxes around them and then labels them accordingly. Photos are crops from EWD images (Schouten et al., 2024) 212
14.3 Diagram of the Faster R-CNN architecture 213
14.4 Data fusion techniques: (left) early fusion, and (right) late fusion 215
14.5 An overview of our multimodal object detection solution 216
14.6 An example of an EWD image sliced into tiles, taken from Schouten et al. (2024). The dashed lines show equally sized tiles. Note the difference in the number of cut wildflowers (Calty palustris in this case) between the two tiling schemes 218
14.7 Selected flower species grouped by visual similarity. Group 1: Buttercup (aggregate), Caltha palustris, Ficaria verna; Group 2: Bellis perennis, Chamomile (aggregate), Leucanthemum vulgare. Photos are crops randomly sampled from EWD 219
14.8 Data alignment overview for both groups: (top) flowering phenology estimates from NDFF for the selected species and (bottom) histogram of objects counts from EWD for the selected species. The horizontal axis is the day of year ranging from 1 to 365, while the vertical axis is (top) the phenological index, normalized from 0 to 1, and (bottom) an object count 220
14.9 Confusion matrix with confidence threshold over 0.75 and IoU threshold over 0.50 for image-only and learned feature-level fusion elementwise addition models 227
15.1 Overview of the end-to-end architecture in ARISE 235
15.2 Manual data collection and identification (top) versus a fully automatic AI-powered solution for monitoring biodiversity (bottom) 237
15.3 Diversity of digital biodiversity sensors tested in ARISE. (a) Location of the three ARISE monitoring demonstration sites in the Netherlands and the deployed sensors to monitor biodiversity non-invasively and remotely. (b) Different sensors and their data volumes 239
15.4 Overview of the Biocloud architecture with the different layers of processing the original data sources from raw to enriched and curated data for future use and access 243
15.5 The active learning cycle of advanced species identification in ARISE 247
Tables
7.1 Examples of changes through technology in various aspects of daily life 103
7.2 Types of Human Nature Connectedness (after Ives et al., 2018) and examples of the role of technology 103
7.3 Ways in which the different types of HNC may be influenced in the three case studies 112
7.4 Future areas of life enhancing HNC 116
7.5 The potential role of technology in the symbiocene 116
10.1 Metrics of green AI 156
14.1 Selected flower species dataset description. Subspecies visually indistinguishable in the field are merged in the EWD dataset: Ranunculus acris and Ranunculus repens are labelled as Buttercup (aggregate), while Matricaria chamomilla and Matricaria maritima are labelled as Chamomile (aggregate) 219
14.2 Test results on Group 1 and Group 2 for all models averaged over five seeds. Best results are shown in bold 224
14.3 Test results per species class for the image-only baseline and the best performing feature fusion models averaged over five seeds 225
14.4 Test results on Group 1 and Group 2 for all classification models averaged over five seeds, hence average precision (AP) 226
15.1 Mapping of processes and scenarios to ARISE components 247