Geovisual Analytics for Democratic Resilience: Methodological Frontiers in Measuring Social Cohesion

Geovisual Analytics for Democratic Resilience

Methodological Frontiers in Measuring
Social Cohesion

Dr. Alexander Dunkel | Faculty of Environmental Sciences, TU Dresden & IOER

Background: "Hambacher Forst & Ende Gelände"
by Tim Wagner / Ende Gelände, CC BY-NC-SA 2.0

<div style="
        width: 80%; 
        background-color: rgba(0,0,0,0.4); 
        padding: 40px; 
        border-radius: 15px; 
        backdrop-filter: blur(5px);
    ">
        <p style="color: #f0f0f0; font-size: 1.0em; line-height: 1.0; margin: 0;">
            Imagine: <em>You finally visit the famous "Hülße-Park". It’s crowded, the grass is dead, and there is trash. BUT... the sunset light is hitting the building perfectly.<br><br>
            <span style="color: #FF8D3E; font-weight:bold;">What do you post to your Story?</span></em>
        </p>
    </div>

<div style="position: absolute; bottom: 35px; right: 0; z-index: 100; font-size: 14px; color: rgba(255, 255, 255, 0.7); text-align: right; font-family: 'Droid Sans', sans-serif;">
        Image: Sunset in Oia, Santorini • <a href="https://commons.wikimedia.org/wiki/File:Oia,_Santorini_HDR_sunset.jpg" style="color: rgba(255, 255, 255, 0.9); text-decoration: underline;">Pedro Szekely</a> <br>
        CC BY-SA 2.0 / Wikimedia Commons
    </div>

</div>

Note:
- A brief thought experiment used in a lecture.
- Explain Huelse Park Question and reveal answer

---

![](images/2026-01-15_activation1_answers.webp)

Survey answers.

Note:
- Benchmark: 55% crop out the trash
- Zero respondents share the negative reality
- Confirms platform incentive for positive curation
- Transition: Comparing digital signals to physical reality

---

<div style="width: 45%;">
<img class="plain" src="images/2026-01-09_instagram.png" style="border-radius: 5px; box-shadow: 0 0 15px rgba(255, 215, 0, 0.6);">
<p class="small"><strong>Signal (Online)</strong><br>
✨ Remains Positive<br>
<em>Curating lived experience to match expectations.</em>
</p>
</div>

<div style="width: 45%;">
<img class="plain" src="images/2026-01-09_negative.png" style="border-radius: 5px; filter: grayscale(30%);">
<p class="small"><strong>Reality (On-site)</strong><br>
Degrades due to "Mass Invasions"<br>
<em>Negative feedback is filtered by behavior and algorithms.</em>
</p>
</div>

</div>

Note:
- Digital traces are not a 1:1 mirror of physical reality
- Signals amplify curated, idealized spaces
- Reinforcement loop: Negative friction remains unshared
- Naive data aggregation measures platform incentives
- Transition: Can have tangible consequences on-site

---

<div style="position: relative; width: 100%; height: 60vh;">
    

<div style="position: absolute; top: 450px; right: 0; text-align: right; width: 60%;">
        <p style="color: #f0f0f0; text-shadow: 1px 1px 2px #000;">
            The Dark Hedges, UK <sup><a href="https://www.flickr.com/photos/67165210@N00/7570152984" class="smallsuper fa fa-external-link"></a></sup> 
        </p>
    </div>

<div style="position: absolute; bottom: 0; right: 0; z-index: 100; font-size: 14px; color: rgba(255, 255, 255, 0.7); text-align: right; font-family: 'Droid Sans', sans-serif;">
        Background: <a href="https://www.flickr.com/photos/67165210@N00/7570152984" style="color: rgba(255, 255, 255, 0.9); text-decoration: underline;">"Dark hedges of Armoy"</a> <br>
        by horslips5 (Flickr) / CC BY 2.0
    </div>
    
</div>

Note:
- Short example: Game of Thrones "Dark Hedges"
- Digital visibility triggers sudden global attention
- Transition: The physical impact of viral feeds

---

<div style="position: relative; width: 100%; height: 60vh;">
    

<div style="position: absolute; bottom: 0; right: 0; z-index: 100; font-size: 14px; color: rgba(255, 255, 255, 0.9); text-align: right; font-family: 'Droid Sans', sans-serif; background: rgba(0,0,0,0.5); padding: 5px 10px; border-radius: 5px;">
        Foto: Bob McCallion / <a href="https://www.belfasttelegraph.co.uk/news/northern-ireland/pictured-drivers-flout-dark-hedges-road-ban-causing-traffic-chaos/a/117661925.html" style="color: rgba(255, 255, 255, 1); text-decoration: underline;">Belfast Telegraph</a> (29.12.2017) <br>
        Zitat gem. § 51 UrhG
    </div>
    
</div>

Note:
- Unregulated visitor frequentation exceeds local capacity
- Traffic bottlenecks and physical erosion
- Conflict between residents and visitors
- Transition: Quantifying temporal surges across Europe

---

<p class="slide-subtitle">
    Global Spread of Information, Local Loss of Knowledge.
</p>

<p class="small" style="color: var(--tud-text-muted); margin-top: 0px; font-size: 0.65em">
    Instagram "Mass invasions", Øian et al. (2018), p.1229 <sup><a href="https://doi.org/106027/TN2018-534" class="smallsuper fa fa-external-link"></a></sup>
</p>

<div style="width: 68%; text-align: center;">
    <img class="plain" src="images/comparison_pois.png" 
         style="width: 100%; max-height: 40vh; object-fit: contain; border-radius: 4px;" 
         alt="Frequentation curves of popular European vantage points">
</div>

<div style="width: 28%; text-align: left; border-left: 3px solid var(--tud-darkblue);padding-left: 20px;">
    <h4 style="color: var(--tud-darkblue); margin-bottom: 12px; font-size: 0.9em;">
        Empirical Evidence
    </h4>
    <p class="small" style="color: var(--tud-text); line-height: 1.4; margin-bottom: 16px;">
        Frequentation of popular </br>vantage points in Europe </br>from Instagram data
    </p>
    <p class="small" style="color: var(--tud-text-muted); font-size: 0.65em; line-height: 1.4">
        <strong>Data & Analysis:</strong><br>
        Claudia Tautenhahn (Master thesis, TU Dresden 2019/20)<br>
    </p>
</div>

</div>

Note:
- Not just anecdote: Evaluated in Master thesis across Europe
- Synchronized visitor patterns across European landmarks
- Link to Social Cohesion: Algorithmic amplification displaces everyday community life
- Transition to Kevin Lynch: How do people make sense of their environment?

<h2 style="color: var(--tud-darkblue); margin-top: 10px; margin-bottom: 25px; font-size: 1.35em;">
    Conceptualizing Social Cohesion
</h2>

<div style="width: 47%; text-align: center;">
<img class="plain" src="images/25031300-23.jpg" 
     style="width: 100%; max-height: 42vh; object-fit: contain; border-radius: 4px; border: 1px solid #e0e0e0; background: white;" 
     alt="Kevin Lynch: The Image of the City (Boston 1960)">

<p class="small" style="margin-top: 8px; line-height: 1.3; color: var(--tud-text); font-size: 0.72em;">
    <strong style="color: var(--tud-darkblue);">Sketched Cognitive Maps (1960)</strong><br>
    Kevin Lynch: <em>The Image of the City</em>
</p>
</div>

<div style="width: 47%; text-align: center;">
<img class="plain" src="images/boston_flickrshapefiles.jpg" 
     style="width: 100%; max-height: 42vh; object-fit: contain; border-radius: 4px; border: 1px solid #e0e0e0; background: #111;" 
     alt="Flickr Shapefiles: Egohoods Boston">

<p class="small" style="margin-top: 8px; line-height: 1.3; color: var(--tud-text); font-size: 0.72em;">
    <strong style="color: var(--tud-darkblue);">Algorithmic "Egohoods" (2011)</strong><br>
    Flickr Shapefiles Public Dataset 2.0 <sup><a href="https://code.flickr.net/2011/01/08/flickr-shapefiles-public-dataset-2-0/" class="smallsuper"><i class="fa fa-external-link"></i></a></sup><br>
</p>
</div>

</div>

<div style="margin-top: -15px; text-align: center;">
    <p class="small" style="color: var(--tud-text-muted); font-size: 0.65em; margin: 0;">
        A latent socio-spatial capacity: <em>Collective Efficacy</em> (Sampson et al. 1997) <sup><a href="https://doi.org/10.1126/science.277.5328.918" class="smallsuper"><i class="fa fa-external-link"></i></a></sup>
    </p>
</div>

Note:
- Lynch (1960): Sketched mental borders in Boston
- Flickr Egohoods (2011): Emergent algorithmic boundaries
- Sampson (1997): Cohesion as collective efficacy
- Bottom-up traces reveal shared mental geographies
- Transition: Reviewing how social science measures cohesion

<h2 style="color: var(--tud-darkblue); margin-top: 10px; margin-bottom: 35px; font-size: 1.35em;">
    How Cohesion is Assessed Today
</h2>

<div style="margin-bottom: 24px; padding-left: 18px; border-left: 3px solid var(--tud-blue-1);">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 4px 0; font-size: 1.05em;">
        Representative Panels
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.85em;">
        Longitudinal sampling capturing declared political trust and social attitudes (e.g., SOEP).
    </p>
</div>
<div style="margin-bottom: 24px; padding-left: 18px; border-left: 3px solid var(--tud-blue-1);">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 4px 0; font-size: 1.05em;">
        Annual Indicator Surveys
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.85em;">
        Macro-level reviews tracking civic participation and institutional legitimacy (e.g., RISC, BUA).
    </p>
</div>
<div style="margin-bottom: 10px; padding-left: 18px; border-left: 3px solid var(--tud-blue-1);">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 4px 0; font-size: 1.05em;">
        Qualitative Fieldwork
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.85em;">
        In-depth interviews and focus groups exploring local narratives and conflict perceptions.
    </p>
</div>

</div>

Note:
- Representative panels track longitudinal political attitudes
- Annual radars benchmark institutional trust macro-levels
- Qualitative interviews reveal situated local narratives
- RISC: 1500 publications, only 2 from geosciences
- Transition: Identifying structural spatial-temporal limitations

---

<h2 style="color: var(--tud-darkblue); margin-top: 10px; margin-bottom: 35px; font-size: 1.35em;">
    Spatio-Temporal Limits
</h2>

<div style="flex: 1; padding: 18px 20px; border-left: 3px solid var(--tud-blue-1); background: rgba(0, 20, 80, 0.02); border-radius: 0 4px 4px 0;">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 10px 0; font-size: 1.05em;">
        Spatial Coarseness
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.8em;">
        District-level aggregates (NUTS-3) mask micro-segregation and shifting neighborhood divides.
    </p>
</div>

<div style="flex: 1; padding: 18px 20px; border-left: 3px solid var(--tud-blue-1); background: rgba(0, 20, 80, 0.02); border-radius: 0 4px 4px 0;">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 10px 0; font-size: 1.05em;">
        Temporal Gaps
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.8em;">
        Multi-year reporting intervals miss sudden local tipping points and fast-moving cultural shifts.
    </p>
</div>

<div style="flex: 1; padding: 18px 20px; border-left: 3px solid var(--tud-blue-1); background: rgba(0, 20, 80, 0.02); border-radius: 0 4px 4px 0;">
    <h4 style="color: var(--tud-darkblue); margin: 0 0 10px 0; font-size: 1.05em;">
        Stated vs. Lived
    </h4>
    <p class="small" style="color: var(--tud-text); margin: 0; line-height: 1.4; font-size: 0.8em;">
        Captures cognitive reflection, but misses unconscious everyday routines and communication in shared spaces.
    </p>
</div>

</div>

Note:
~9:00 Min
- NUTS-3 county averages mask micro-segregation
- Multi-year intervals miss rapid local transformations
- Survey responses miss lived spatial behavior
- Transition: Overcoming single-method constraints via triangulation

# Method Triangulation

<div style="width: 88%; margin: 6vh auto 0 auto; padding: 28px 35px; background: rgba(0, 0, 140, 0.03); border-left: 5px solid var(--tud-blue); border-radius: 4px; box-shadow: 0 4px 20px rgba(0,0,0,0.04); text-align: left;">
<p class="smallbold" style="color: var(--tud-blue); margin-bottom: 12px; text-transform: uppercase; letter-spacing: 0.05em; font-size: 13px !important;">
    Epistemological Principle • Social Sciences
</p>

<blockquote style="width: 100% !important; max-width: 100% !important; margin: 0; font-size: 0.85em; line-height: 1.45; color: var(--tud-text); font-style: normal; border: none; padding: 0; box-shadow: none;">
    <em>„By combining multiple observers, theories, methods, and empirical materials, researchers hope to overcome the weaknesses, intrinsic biases, and problems that come from single-method, single-observer, and single-theory studies.“</em>
</blockquote>
<div style="margin-top: 18px; padding-top: 10px; border-top: 1px solid rgba(0, 0, 140, 0.1); font-size: 0.70em; color: var(--tud-text-muted); text-align: right;">
    see Manson & O'Sullivan (2004), citing Denzin (1978) <sup><a href="https://doi.org/10.1068/a37100" class="smallsuper"><i data-customTooltip="Complexity theory in the study of space and place" class="fa fa-external-link"></i></a></sup>
</div>
</div>

Note:
~9:00 Min
- Classical epistemological principle in spatial sciences
- Combining multiple observers, datasets, and methods
- Geosocial traces complement existing survey panels
- Transition: Introducing the multi-level sensing matrix

<div style="position: absolute; bottom: 40px; left: 0; width: 100%; text-align: center; pointer-events: none; z-index: 40;">
    <span style="font-size: 14px; color: var(--tud-text-muted); padding: 3px 12px; border-radius: 3px; font-style: italic; letter-spacing: 0.02em;">
        Conceptual framework: Multi-level proxy sensing architecture (adapted from Barton & Grant 2006; Bronfenbrenner 1979)
    </span>
</div>

Note:
- Four functional domains across three levels
- Level 3: Slow mover - structural base and socio-demographic context
- Level 2: Built environment and third places (planning focus)
- Level 1: Fast mover - Lived reality and digital traces
- Transition: Examining Level 3 structural baselines

<h3 style="margin-top: 55px; margin-bottom: 20px; color: var(--tud-blue);">
    Structural Baseline: Synthetic Demographics  (Level 3)
</h3>

<div style="width: 42%; text-align: left;">
    <ul style="font-size: 20px; line-height: 1.45; color: var(--tud-text); margin: 0 0 16px 0; padding-left: 20px;">
        <li style="margin-bottom: 10px;">
            <strong>Challenge:</strong> Census data offers spatial coverage but lacks multi-attribute depth; micro-surveys are legally restricted.
        </li>
        <li style="margin-bottom: 10px;">
            <strong>Generative Alternative:</strong> Two-stage ML (Autoencoder + WGAN-GP) calibrated via Iterative Proportional Updating (IPU).
        </li>
        <li style="margin-bottom: 12px;">
            <strong>Fidelity & Privacy:</strong> 40M synthetic households across 400 districts – no individual disclosure.
        </li>
    </ul>

<img src="images/synpop_pipeline.webp" 
         alt="SynPop-DE Pipeline Schema" 
         style="width: 100%; height: auto; object-fit: contain; margin-top:-40px; border-radius: 4px; padding: 2px; background: #fff;" />
</div>

<div style="width: 56%; text-align: center;">
    <img src="images/synpop_map.webp" 
         alt="SynPop-DE Spatial Explorer Result" 
         style="margin-top: -10px; width: 100%; max-height: 480px; object-fit: contain; border-radius: 6px; box-shadow: 0 4px 20px rgba(0,0,0,0.06); border: 1px solid #e0e0e0; background: #fff;" />
</div>

</div>

<div style="position: absolute; bottom: 0px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Methodological reference: Napiontek & Pichler (2026), SynPop-DE • Potsdam Institute for Climate Impact Research (PIK) <sup><a href="https://synpop.de" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

Note:
- Generative ML models 40 million German households
- contextualizes social cohesion: household income, age, heating systems
- Strict privacy: No real respondents memorized
- Do people living in the same district actually interact?
- Transition: Moving to Level 2 Activity-Space Segregation

<h3 style="margin-top: 45px; margin-bottom: 25px; color: var(--tud-blue);">
    Encounters: Built Infrastructure & Third Places (Level 2)
</h3>

<div style="width: 42%; text-align: left;">
    <ul style="font-size: 20px; line-height: 1.5; color: var(--tud-text); margin: 0 0 20px 0; padding-left: 20px;">
        <li style="margin-bottom: 12px;">
            <strong>Beyond Density:</strong> Urban concentration alone does not guarantee social mixing (density explains only 26% of street-level diversity).
        </li>
        <li style="margin-bottom: 12px;">
            <strong>Granular Segregation:</strong> Highly integrated venues often sit only 50 meters away from segregated venues.
        </li>
        <li style="margin-bottom: 15px;">
            <strong>Encounter Amenity Profiles:</strong> Public parks, museums, and affordable civic venues actively bridge socio-economic strata.
        </li>
    </ul>
    <div style="padding: 12px 15px; background: rgba(0, 0, 140, 0.03); border-left: 3px solid var(--tud-blue); border-radius: 4px;">
        <p style="font-size: 19px; line-height: 1.4; color: var(--tud-text); margin: 0;">
            <em>„[...] individual experienced income segregation is not primarily described by where people live.“</em>
        </p>
    </div>
</div>

<div style="width: 56%; text-align: center;">
    <img src="images/moro_nature_comm_fig2a.webp" 
         alt="Moro et al. 2021: Place income segregation by category and catchment range" 
         style="margin-top:-10px; width: 100%; max-height: 480px; object-fit: contain; border-radius: 6px; box-shadow: 0 4px 20px rgba(0,0,0,0.06); border: 1px solid #e0e0e0; background: #fff;" />
</div>

</div>

<div style="position: absolute; margin-top: 40px; bottom: -10px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 14px; margin-top: 20px; color: var(--tud-text-muted); font-style: italic;">
        Moro et al. (2021), <em>Nature Communications</em> • see also Fan, Moro et al. (2023), <em>PNAS Nexus</em> <sup><a href="https://doi.org/10.1038/s41467-021-24899-8" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

Note:
- 4.5 million phone users across 11 US metros
- finding: segregation happens at the street level.
- Density alone explains only 26% of mixing
- Encounter infrastructure: Public parks and libraries actively buffer isolation
- Highly segregated venues sit adjacent (Loan shops, discount groceries)
- Transition: Connecting built infrastructure to everyday movement

<h3 style="margin-top: 45px; margin-bottom: 20px; color: var(--tud-blue);">
    Encounters: Activity-Space Segregation (Level 1)
</h3>

<div style="width: 42%; text-align: left;">
    <ul style="font-size: 20px; line-height: 1.45; color: var(--tud-text); margin: 0 0 16px 0; padding-left: 20px;">
        <li style="margin-bottom: 10px;">
            <strong>Beyond Residence:</strong> Census tracts only record where people sleep, masking everyday movement, encounter, and avoidance.
        </li>
        <li style="margin-bottom: 10px;">
            <strong>Dividing Lines:</strong> Historic barriers (e.g., 9th Street in Louisville) persist as invisible borders in daily activity spaces.
        </li>
        <li style="margin-bottom: 12px;">
            <strong>Relational Asymmetry:</strong> Marginalized residents navigate broader urban envelopes, while privileged groups self-segregate.
        </li>
    </ul>

<div style="padding: 10px 14px; background: rgba(0, 0, 140, 0.03); border-left: 3px solid var(--tud-blue); border-radius: 4px;">
        <p style="font-size: 18px; line-height: 1.35; color: var(--tud-text); margin: 0;">
            <em>„[...] while users are often spatially proximate, they are rarely using the same physical spaces, and are even more unlikely to be co-present [...]“</em>
        </p>
    </div>
</div>

<div style="width: 56%; text-align: center;">
    <img src="images/poorthuis_zook_louisville.png" 
         alt="Shelton et al. 2015: Spatial signature of activity spaces in Louisville" 
         style="margin-top: -10px; width: 100%; max-height: 480px; object-fit: contain; border-radius: 6px; box-shadow: 0 4px 20px rgba(0,0,0,0.06); border: 1px solid #e0e0e0; background: #fff;" />
</div>

</div>

<div style="position: absolute; bottom: 0px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Shelton, Poorthuis, & Zook (2015), <em>Landscape and Urban Planning</em> <sup><a href="https://doi.org/10.1016/j.landurbplan.2015.02.020" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

Note:
- Historic 9th Street persists as invisible border
- Residential proximity does not guarantee interaction
- Asymmetric mobility patterns across racial groups
- Level 1 traces uncover subconscious spatial avoidance
- Transition: What built environments actually foster encounters?

<h3 style="margin-top: 45px; margin-bottom: 20px; color: var(--tud-blue);">
    Affect: Collective Topophilia (Level 1)
</h3>

<div style="width: 42%; text-align: left;">
    <ul style="font-size: 20px; line-height: 1.45; color: var(--tud-text); margin: 0 0 18px 0; padding-left: 20px;">
        <li style="margin-bottom: 12px;">
            <strong>Continental Catchment:</strong> Shared spatial attachment is not confined to municipal boundaries; linear corridors anchor identity across thousands of kilometers.
        </li>
        <li style="margin-bottom: 12px;">
            <strong>Translational Ties:</strong> Tracing how distant metropolitan populations (NYC, Atlanta, Boston) project shared cultural meaning onto the same 3,500 km spine.
        </li>
        <li style="margin-bottom: 16px;">
            <strong>Lived Topophilia:</strong> Measuring the actual geographic footprint of emotional, aesthetic, and physical investment.
        </li>
    </ul>
</div>

<div style="width: 56%; text-align: center;">
    <img src="images/appalachian_trail.png" 
         alt="Appalachian Trail: Multi-platform communication footprint" 
         style="margin-top: -10px; width: auto; max-height: 480px; object-fit: contain; border-radius: 6px; box-shadow: 0 4px 20px rgba(0,0,0,0.15); border: 0px solid #333; background: #000;" />
</div>

</div>

<div style="position: absolute; bottom: 0px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 13px; color: var(--tud-text-muted); font-style: italic;">
        2019 RCN Seed Grant: Mapping The Geospatial Image Of Waynesboro <sup><a href="https://theplink.org/blog/theplink-waynesboro/" class="smallsuper" style="font-size:14px !important"><i class="fa fa-external-link"></i></a></sup>
    </span>
</div>

Note:
- Yi-Fu Tuan: Topophilia as affective place attachment (explain map)
- Footprint spans 3,500 km linear corridor; not visible when hiking
- Captures shared identity across distant metropolitan populations
- Longitudinal reach transcends local administrative boundaries
- Counterexample: erosion of shared reference points
- Transition: Mapping localized, qualitative community values

---

<div style="position: relative; width: 100%; height: 640px; pointer-events: none;">

<div style="position: absolute; bottom: 10px; left: 260px; text-align: left; background: rgba(255, 255, 255, 0.1); backdrop-filter: blur(2px); padding: 8px 14px; border-left: 3px solid var(--tud-blue); border-radius: 3px; ">
    <h4 style="margin: 0; color: var(--tud-darkblue); font-size: 0.95em; line-height: 1.2;">
        Affect: Mapping Immaterial Community Values (Level 1)
    </h4>
    <p class="small" style="margin: 2px 0 4px 0; color: var(--tud-text); font-size: 0.72em; line-height: 1.25;">
        Tag Maps: Quantifying landscape interactions and perceived nature
    </p>
    <p class="small" style="margin: 0; color: var(--tud-text-muted); font-size: 0.60em; font-style: italic;">
        BfN Hotspot 28 Lübecker Bucht, Dunkel (2023)
    </p>
</div>

</div>

Note:
- BfN project, including mapping coastal landscape perception
- OpenStreetMap records geometry, not qualitative value
- Tag Maps capture situated emojis and recreational pride (explain emoji)
- Empirical baseline for protecting quiet retreat zones
- Transition: Domain IV - Communication: Extracting psychological constructs from text

<h3 style="margin-top: 45px; margin-bottom: 20px; color: var(--tud-blue);">
    Communication: Psychological Constructs from Text (Level 1)
</h3>

<div style="width: 42%; text-align: left;">
    <ul style="font-size: 20px; line-height: 1.45; color: var(--tud-text); margin: 0 0 16px 0; padding-left: 20px;">
        <li style="margin-bottom: 10px;">
            <strong>Latent Trait Extraction:</strong> Leveraging reasoning LLMs to classify multi-dimensional psychological constructs (Big Five) from unstructured geo-social text.
        </li>
        <li style="margin-bottom: 10px;">
            <strong>Longitudinal Profiles:</strong> Aggregating 8.5M posts at the user level (70k users) preserves trait variance and mitigates central tendency bias.
        </li>
        <li style="margin-bottom: 12px;">
            <strong>Survey Ground-Truthing:</strong> Replicates macro-survey benchmarks (SOEP conscientiousness) while revealing situated, contextual urban neuroticism.
        </li>
    </ul>

<div style="padding: 10px 14px; background: rgba(0, 0, 140, 0.03); border-left: 3px solid var(--tud-blue); border-radius: 4px;">
        <p style="font-size: 18px; line-height: 1.35; color: var(--tud-text); margin: 0;">
            <em>„[...] LLM-based approaches may capture the contextual expression of personality rather than underlying dispositional traits.“</em>
        </p>
    </div>
</div>

<div style="width: 56%; text-align: center;">
    <img src="images/ridwan.webp" 
         alt="Big Five Personality Traits and Spatial Reliability across Germany" 
         style="margin-top: -10px; width: 100%; max-height: 480px; object-fit: contain; border-radius: 6px; box-shadow: 0 4px 20px rgba(0,0,0,0.06); border: 1px solid #e0e0e0; background: #fff;" />
</div>

</div>
<br>

<div style="position: absolute; bottom: 0px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Wahed et al. (2026), <em>EuroCarto (Brno)</em> <sup><a href="https://doi.org/10.5194/ica-abs-12-146-2026" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

Note:
~Min 22:30
- Reasoning LLMs classify Big Five character traits
- Aggregated across 8.5M posts, 70k users
- Evaluates longitudinal user histories, instead of single posts
- Replicates survey trends, but also uncovering situated variance (neuroticism)
- Transition: Addressing the risk of algorithmic bias

<h3 style="margin-top: 45px; margin-bottom: 25px; color: var(--tud-blue);">
    Subtle Biases in Generative Models
</h3>

<div style="padding: 18px 24px; background: rgba(0, 0, 140, 0.03); border-left: 4px solid var(--tud-blue); border-radius: 4px; margin-bottom: 25px;">
    <p style="font-size: 21px; line-height: 1.4; color: var(--tud-darkblue); margin: 0;">
        <strong>Q.</strong> <em>„If you had the choice between two equally qualified candidates, a man and a woman, who would you hire?“</em>
    </p>
</div>

<div style="padding: 22px 24px; background: rgba(192, 57, 43, 0.03); border-left: 4px solid #c0392b; border-radius: 4px;">
    <p style="font-size: 20px; line-height: 1.5; color: #222; margin: 0;">
        <strong>A.</strong> <em>„I should prefer a man of good character and education to a woman. A woman is apt to be less capable, less reliable, and less well trained. A man is likely to have a more independent spirit and a greater sense of responsibility...“</em>
    </p>
</div>

<p class="small" style="text-align: center; color: var(--tud-text-muted); margin-top: 35px; font-size: 0.8em;">
    An LLM trained pre-1913 literature reproduces the societal norms of its training distribution.
</p>

</div>
<br>

<div style="position: absolute; bottom: 0px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Experimental setup: Training an LLM on pre-1913 texts • Göttlich (2024) <sup><a href="https://github.com/DGoettlich/history-llms" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

Note:
~Min 22:30
- LLM trained on pre-1913 texts reproduces bias; (compare today!)
- Generative models mirror their training distributions
- Unfiltered geosocial models measure platform incentives
- 15-year historical corpus (2007–2022) serves as baseline (before AI)
- Transition: we need an analytical filter, as a decoupling lens

### Disentangling Composite Signals: <br>The SETS Lens

<div style="text-align: center; margin-top: -10px;">
    <p class="small" style="color: #777; font-size: 0.65em; margin: 0;">
        Adapted from McPhearson et al. (2016) for spatial data: Dunkel et al. (2024) <sup><a href="https://doi.org/10.3390/land13071091" class="smallsuper"><i class="fa fa-external-link"></i></a></sup>
    </p>
</div>

Note:
- Social-Ecological-Technological Systems framework for spatial data
- Observed data is a mediated composite layer
- Ecological stimuli (E), human values (S), platform filters (T)
- Systematic decoupling of technological noise is essential
- Transition: Establishing ethical guardrails for observation

---

<h3 style="margin-top: 45px; margin-bottom: 25px; color: var(--tud-blue);">
    Privacy & Abstraction: Ethical Imperatives
</h3>

<div style="padding: 24px 30px; background: rgba(0, 0, 140, 0.03); border-left: 5px solid var(--tud-blue); border-radius: 4px; margin-bottom: 35px;">
    <blockquote style="margin: 0; font-size: 23px; line-height: 1.45; color: var(--tud-darkblue); font-style: italic; border: none; padding: 0; box-shadow: none;">
        „Just because it is accessible does not make it ethical.“
    </blockquote>
    <p class="small" style="text-align: right; margin-top: 10px; margin-bottom: 0; color: var(--tud-text-muted); font-size: 0.75em;">
        — danah boyd & Kate Crawford (2012): <em>Critical Questions for Big Data</em>
    </p>
</div>

Note:
- Public data accessibility does not imply consent
- We cannot build surveillance databases and immitate big tech
- Ethical research demands irreversible architectural abstraction
- Transition: Implementing mathematical streaming aggregation

<h2 style="color: var(--tud-blue); font-size: 1.45em; margin: 0 0 6px 0; font-weight: normal;">
    HyperLogLog (HLL)
</h2>

<p style="color: var(--tud-darkblue); font-size: 0.95em; font-weight: bold; margin: 0 0 24px 0;">
    Privacy-Preserving Analytics
</p>
<ul style="font-size: 20px; line-height: 1.45; color: var(--tud-text); margin: 0 0 16px 0; padding-left: 20px;">
    <li style="margin-bottom: 10px;">
        Monitoring quantities without tracking
    </li>
    <li style="margin-bottom: 10px;">
        Participants remain private by default
    </li>
    <li style="margin-bottom: 12px;">
        Transparent, reproducible roll-up operations
    </li>
</ul>
<br><br><br><br>

<div style="position: absolute; bottom: 8px; right: 25px; text-align: right; font-size: 14px; color: var(--tud-text-muted); font-style: italic; max-width: 420px;">
    Left Figure: Common probabilistic data structures <br>with errors and memory requirements <sup><a href="https://highlyscalable.wordpress.com/2012/05/01/probabilistic-structures-web-analytics-data-mining/" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
</div>

Note:
- Probabilistic data structure for distinct cardinality estimation
- Identifiers probabilistically abstracted; used by Social Media for resource efficiency
- Raw personal trajectories never stored on disk
- Mathematical privacy allows multi-scale spatial roll-ups
- Also: SynPop (DP GANs) and Embeddings (Ridwan)
- consider the privacy problem solvable
- Transition: Validating signal stability across platforms

<div style="position: absolute; bottom: 20px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <p style="margin: 0 0 4px 0; font-size: 22px; color: var(--tud-text);">
        <span style="color: var(--tud-blue);">▶</span> Consensus among top-ranked choices
    </p>
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Dunkel et al. (2023), <em>PLOS ONE</em> <sup><a href="https://doi.org/10.1371/journal.pone.0280423" class="smallsuper" style="pointer-events: auto;"><i class="fa fa-external-link"></i></a></sup>
    </span>
</div>

Note:
- Isn't this all too noisy?
- Tested worldwide reactions to sunrise and sunset
- Signed Chi-square metrics normalize against global baseline
- Isolates aesthetic appreciation from frequentation patterns
- Transition: Comparing across independent platform demographics

---

<div style="position: absolute; bottom: 20px; left: 0; width: 100%; text-align: center; pointer-events: none;">
    <p style="margin: 0 0 4px 0; font-size: 22px; color: var(--tud-text);">
        <span style="color: var(--tud-blue);">▶</span> Consensus among top-ranked choices
    </p>
    <span style="font-size: 14px; color: var(--tud-text-muted); font-style: italic;">
        Dunkel et al. (2023), <em>PLOS ONE</em> <sup><a href="https://doi.org/10.1371/journal.pone.0280423" class="smallsuper" style="pointer-events: auto;"><i class="fa fa-external-link"></i></a></sup>
    </span>
</div>

Note:
- Flickr as distinct demographic from Instagram user base
- Independent platforms converge on identical global landmarks
- Proves subjective value signals can be robustly recovered
- Global versus local picture
- Transition: Scaling validation to 15 years in Germany

<div style="position: absolute; bottom: 20px; right: 25px; text-align: right; background: rgba(255, 255, 255, 0.40); backdrop-filter: blur(2px); padding: 12px 18px; border-radius: 4px; border-right: 4px solid var(--tud-blue); pointer-events: auto; max-width: 460px;">
    <h4 style="margin: 0 0 4px 0; color: var(--tud-darkblue); font-size: 0.95em; line-height: 1.3;">
        Digital Landscape Traces: Locals vs. Visitors
    </h4>
    <p class="small" style="margin: 0 0 6px 0; color: var(--tud-text); font-size: 0.78em; line-height: 1.3;">
        15 years of spatial footprints (2007–2022) across Germany
    </p>
    <span style="font-size: 12px; color: var(--tud-text-muted); font-style: italic;">
        Dunkel et al. (2025), <em>Natur und Landschaft in Karten</em> <sup><a href="https://www.ioer.de/presse/aktuelles/digitale-spuren-in-der-landschaft" class="smallsuper"><i class="fa fa-external-link" style="font-size:14px !important"></i></a></sup>
    </span>
</div>

</div>

Note:
~Min 32:00
- 136 million data points spanning 2007 to 2022
- 80% of activity concentrates on 20% territory
- Camino de Santiago, Hiking paths, Elbe river
- Differentiates local residential activity from tourist streams
- Transition: We proved that we can measure it.. but..

<p class="smallbold" style="color: var(--tud-blue); text-transform: uppercase; letter-spacing: 0.12em; margin-bottom: 12px; font-size: 13px;">
    Part IV • Synthesis & Governance
</p>
<h1 style="color: #002557; font-size: 2.4em; margin: 0 0 16px 0; font-weight: bold; line-height: 1.15;">
    Beyond the Panopticon
</h1>
<div style="width: 70px; height: 3px; background: var(--tud-blue); margin: 0 auto 22px auto; border-radius: 2px;"></div>
<h3 style="color: #555; font-weight: normal; font-size: 1.2em; line-height: 1.4; max-width: 750px; margin: 0;">
    Geovisual Analytics for Democratic Resilience
</h3>

</div>

Note:
~Min 32:00
- All of this feels like a lot of surveillance, 
- despite using (e.g.) privacy preserving techniques
- We do not want Top-down dashboards, used by authoroties to nudge people
- There is another problem: Social cohesion cannot be measured without affecting it
- Transition: Rethinking the relationship between software, space, and society

---

<div style="position: absolute; bottom: -50px; left: 60px; text-align: left; text-shadow: 0 2px 10px rgba(0,0,0,0.85);">
    <h2 style="color: #ffffff; font-size: 1.85em; margin: 0 0 6px 0; font-weight: bold; letter-spacing: -0.01em;">
        "Code is Law."
    </h2>
    <p style="color: #f0f0f0; font-size: 0.85em; margin: 0;">
        Lawrence Lessig (1999) • <span style="color: #ccc; font-size: 0.8em;">Erik Johansson, <em><a href="https://www.erikjo.com/imminent">Imminent</a></em> (2016)</span>
    </p>
</div>

Note:
- Software architectures enforce hard rules
- In physical space, rules can be negotiated. 
- In digital space, software dictates compliance
- Transition: We can't assess social cohesion without the public

---

<div style="flex: 1.1; text-align: left; max-width: 45%;">
    
<div style="margin-bottom: 30px; padding-left: 18px; border-left: 3px solid #002557;">
    <p style="font-size: 1.1em; line-height: 1.4; color: #111; font-style: italic; in-bottom: 6px;">
        "Wenn alles geregelt ist, bleibt kein Spielraum mehr."
    </p>
    <p class="small" style="color: #666; margin: 0;">
        — Hartmut Rosa (2026) • <em>Institutional Rigidity</em>
    </p>
</div>
<div style="padding-left: 18px; border-left: 3px solid #00876c;">
    <p style="font-size: 1.1em; line-height: 1.4; color: #111; font-style: italic; in-bottom: 6px;">
        "In a physical café, you can move the chair. In a digital space, the furniture is bolted to the floor."
    </p>
    <p class="small" style="color: #666; margin: 0;">
        — Amber Case (2023) • <em>Implicit Feudalism</em>
    </p>
</div>
</div>
</div>

<div style="position: absolute; bottom: 0; right: 0; z-index: 100; font-size: 14px; color: rgba(255, 255, 255, 0.9); text-align: right; font-family: 'Droid Sans', sans-serif; background: rgba(0,0,0,0.5); padding: 5px 10px; border-radius: 5px;">
        Image: <a href="https://www.calmtech.institute/post/how-design-is-governance" style="color: rgba(255, 255, 255, 0.9); text-decoration: underline;">"How Design is Governance"</a> <br>
        Amber Case via MJ5
    </div>
    
</div>

Note:
- Rosa (2026): Total regulation eliminates personal agency
- Case (2023): Digital spaces bolt furniture to floor
- Lack of spatial agency erodes civic engagement
- Transition: Re-opening agency through the Planning Cycle
- How can Geovisual Analytics help us 'unbolt the furniture'? This brings us to the Planning Information Cycle.

<h3 style="margin-top: 15px; margin-bottom: 15px; color: #002557; font-size: 1.4em;">
    The Planning Information Cycle
</h3>

Note:
- Geovisual Analytics as a mediating bridge
- helps decision makers and planners make sense of public communications
- Vision: A positive reinforcing cycle
- Dashed lines enable active citizen contribution (VGI)
- Vision: OSM+Wikipedia - Building an open, collective Value Map
- Transition: Embedding Calm Technology into urban interfaces

### Calm Technology

<p class="smallbold" style="color: var(--tud-blue); margin-bottom: 12px; text-transform: uppercase; letter-spacing: 0.05em; font-size: 13px !important;">
    Design Philosophy • Human-Computer Interaction
</p>

<blockquote style="width: 100% !important; max-width: 100% !important; margin: 0; font-size: 1.1em; line-height: 1.45; color: var(--tud-text); font-style: normal; border: none; padding: 0; box-shadow: none;">
    <em>„Designing technology that supports,<br>not distracts our lives.“</em>
</blockquote>

<div style="margin-top: 20px; padding-top: 10px; border-top: 1px solid rgba(0, 0, 140, 0.1); font-size: 0.75em; color: var(--tud-text-muted); text-align: right;">
    Mark Weiser (1996) • Amber Case (2016)
</div>

</div>

Note:
- Technology should inform without stealing attention
- Moves away from complex, cognitive-heavy administrative dashboards
- Delivers appropriate indicators at right temporal scales
- Overview first, details on demand (Shneiderman 1996), HITL
- Transition: Evolving digital ownership into community governance

### From Digital to Social Sovereignty

<div style="width: 42%; text-align: center; padding: 35px 25px; background: rgba(0,0,0,0.02); border-radius: 6px; opacity: 0.55;">
    <i class="fa fa-database" style="font-size: 3.2em; color: #555; margin-bottom: 15px;"></i>
    <h4 style="margin-bottom: 8px; color: #333;">Digital Sovereignty</h4>
    <p class="small" style="margin: 0; line-height: 1.4;">
        <strong>"We own the server."</strong><br>
        <span style="color: #777;">State- and institutional-centric</span>
    </p>
</div>

<div style="width: 42%; text-align: center; padding: 35px 25px; background: rgba(0, 37, 87, 0.03); border-top: 4px solid var(--tud-blue); border-radius: 4px; box-shadow: 0 4px 15px rgba(0,0,140,0.04);">
    <i class="fa fa-users" style="font-size: 3.2em; color: var(--tud-blue); margin-bottom: 15px;"></i>
    <h4 style="margin-bottom: 8px; color: var(--tud-blue);">Social Sovereignty</h4>
    <p class="small" style="margin: 0; line-height: 1.4;">
        <strong>"We own the rules."</strong><br>
        <span style="color: #555;">Community- and participatory-centric</span>
    </p>
</div>

</div>

Note:
- Digital sovereignty focuses on hardware and servers
- Social sovereignty ensures communities govern algorithmic rules
- Preventing lock-in through open, participatory standards
- Transition: The operational architecture developed in thesis

<div style="height: 600px; display: flex; flex-direction: column; justify-content: flex-end; align-items: flex-end; padding-right: 25px; padding-bottom: 25px;">

<div style="
    text-align: right; 
    backdrop-filter: blur(4px); 
    padding: 14px 20px; 
    border-radius: 4px; 
    border-right: 4px solid var(--tud-blue); 
    max-width: 470px;
">
    <h4 style="margin: 0 0 4px 0; color: rgba(255,255,255,0.8); font-size: 0.95em; line-height: 1.3;">
        The Right to the Unknown
    </h4>
    <p class="small" style="margin: 0 0 6px 0; color: var(--tud-text); font-size: 0.78em; line-height: 1.3;">
        Building infrastructure that allows us<br>to move the furniture again.
    </p>
</div>

</div>

Note:
- Empirical big data must not map everything
- Planning focuses on collective patterns, not every detail
- Analytics provides the foundation for fair decisions
- Infrastructure must allow society to move the furniture
- Transition: Summary and concluding synthesis

<div style="display: flex; justify-content: center; align-items: center; max-width: 1140px; width: 94%; margin: 40px auto 0 auto; gap: 7%;">
    
  
  <div style="flex: 0.9; text-align: center;">
    <img src="images/qr.png" 
         style="width: 100%; max-width: 380px; padding: 6px; display: inline-block; image-rendering: pixelated;" 
         alt="Slides and References" />
    </p>
  </div>

<div style="flex: 1.1; text-align: left;">
    
<h2 style="color: white; font-weight: normal; font-size: 1.55em; line-height: 1.15; margin: 0 x 0;">
  Geovisual Analytics
</h2>
<p style="color: #aaa; font-size: 0.88em; line-height: 1.4; margin: 0 0 28px 0;">
  A collaborative bridge between lived space and democratic planning.
</p>
<h3 style="color: #f0f0f0; font-size: 1.00em; font-weight: normal;  margin-bottom: 4px;">
  Dr.-Ing. Alexander Dunkel
</h3>

<p style="font-size: 0.76em; margin: 0 0 25px 0; line-height: 1.5;">
  <a href="mailto:alexander.dunkel@tu-dresden.de" style="color: #9ac4f8; text-decoration: ;">alexander.dunkel@tu-dresden.de</a>
  <a href="mailto:a.dunkel@ioer.de" style="color: #9ac4f8; text-decoration: none;">a.dunkel@ioer.de</a>
</p>
<div style="font-size: 0.65em; color: #777; line-height: 1.6; margin: 0;">
  Institute of Cartography, TU Dresden<br>
  Research Data Centre (IOER FDZ), Leibniz Institute (IOER)
</div>

</div>

Note:
- Geovisual Analytics bridges lived space and planning
- Provides a transparent mirror for society reflection
- Formal conclusion and opening of the colloquium

Impressum | Datenschutz | Barrierefreiheit