Computers, Society, and Nature

This conference paper discusses about the spatial-temporal scaling behavior of human mobility by conducting an experimental study using five datasets from different areas and generations. They do get the results that consistent with the literature that human mobility shows characteristics forms for power law distributions, not all datasets are equal, etc. However, what are the basic disciplines of human mobility is not well explained. And the analysis (case studies) carried out based on large amounts of data generated by new measurement techniques is to examine the impact on aggerate metrics of spatial and temporal sampling period. These analysis and discussion on results conduct great researches on how scaling behavior varies and how massive datasets be interpolated with a generalized conclusion that spatial temporal resolution behaviors matter a lot to describe human mobility. That issue is not only associated with human mobility analysis, and it does always matter in plenty of fields in GISicence.

What is particular different and influential on human mobility? Is there any spatial data quality or spatial data uncertainty discussion necessary before or after analyzing movement datasets? Is there any argument on definition of human mobility and related measuring metrics? I expected more about more fundamental issues on human movement analysis which is still vague to me instead of case studies showing basic rules form human mobility and relationships with scaling issues.

Quale — a new word I will almost certainly never use.

It does however represent a concept we all have to wrangle with. Forget statistical models, literally no representation is complete. I tell you to imagine a red house – you imagine it. But was it red? Or maroon, burgundy, pink, or orangish? Not just a matter of precision, what we are communicating depends on what we both think of as ‘red’, or ‘maroon’, or ‘burgundy’, or whatever else. We might also have ideas about what sorts of red are ‘house’ appropriate. An upper-level ontology might suggest to us red-ness that is universal. But no houses in my neighbourhood are the bright lego red? Why not?

Some of what Schade writes reminds me of Intro Stats: error being present in every single observation. This sort of error can be thought of as explained and unexplained variance. Variance is present in all data; the unexplained variety being what may have risen due not only apparatus error, but also what we describe as uncertainty in data.

Schades temperature example is handy: the thermometer doesn’t read 14 degrees – it reads 13.5-14.5 with 68% probability. The stories we tell aren’t about what we say, but what we mean. This sort of anti-reductionism is also at the root of complexity theory. Acknowledging that we cannot characterize systems as static and linear components disregards the emergence of something that can explain why complex things are greater than the sum of their parts. Applied machine learning research also appreciates this anti-reductionism: the link to AI Schade makes, I THINK is about how applied machine learning researchers aren’t really interested in determining the underlying relationships of phenomena – only the observed patterns of associations. Methods that neglect the former and embrace the latter perspective explicitly consider their data to be incomplete and uncertain to some degree. Though to be honest this connection seems forced in the paper, but I’m happy to help force it along. 🙂

This chapter in the book “Fundamentals of Spatial Data Quality” gives a shot on basic concepts in spatial data quality by pointing out that the divergences between the reality and the representation are what the spatial data quality issues often deal with. And there are several aspects of where the errors would happen during the data production process such as the data manipulation process and the human involved data creation process. Moreover, spatial data quality is summarized to be assessed from internal and external aspects. This chapter explains well what the data quality is and what errors could be and is very easy to understand.

It is interesting that the introduction starts with a quote, “All models are wrong, but some are useful”. However, does it mean all spatial data or data created could be interpolated as the product of model or filter? Authors argue that the representation of reality may not be fully detailed and accurate but partially useful. But how to determine whether the data with those uncertainty or errors should be accepted is a much more urgent problem. Also, as the topic is “spatial data uncertainty” and spatial data quality issues discussed in the chapter, does the uncertainty exactly mean different sources of error assessed in spatial data quality?

The chapter defines the internal quality as level of similarity between data produced and perfect data while external quality means level of concordance between data product and user needs. My thought is if user participate in the data producing process (which is about internal quality), will the external quality be efficiently and effectively improved? Can we just replace “as requested by the manager” with “what user wanted” in Figure 2.4 and there should be no external quality worries?

This paper written by Stan Openshaw in 1992 introduces concept of artificial intelligence application in GIS to us by explain how AI emerged and being applied in geographic information system development and why AI is inevitably needed and matters a lot in spatial modelling and analysis. AI does bring a lot to GIS development concerning large spatial database management, spatial data pattern recognition and modifying spatial statistics analysis. Neurocomputing, as a revolution of the century, makes it possible for large data sets analysis and modelling, both supervised classification and unsupervised classification eliminate the difficulties and uncertainty of manual analysis and computing for pattern studies. And AI is definitely unavoidable to be referred to when applying spatial data mining to study large spatial data sets. The paper has a clear structure and explain well the complicated concept of AI application in GIS with a strong background of how and why AI should be used in GIS though I do not fully understand the specific method like expert system and ANN.

As we all known, spatial data in GIS is quite different from general type of data with characteristics of spatial dependencies, space- time dependencies, non-stationarities, etc. and AI technologies give more chance to deal with those complex properties. However, I am wondering if these characteristics in spatial data sets and special way in treating them using AI help develop Ai technologies itself (method structure, algorithm development). Will GIS bring opportunities and development for AI? What GIS have brought for AI?

This article gives us an overview of the importance of research on Geo-visualization topic and discusses some major themes for Geo-visualization and related issues, raising up the main discussion about current challenges emerged in Geo-visualization. Moreover, the authors summarize research challenges and problems proposed crosscutting and end with recommended action for these emerging challenges. Generally speaking, the paper goes through most of research challenges for Geo-visualization from various aspects and lists them one by one for each theme, but it seems not so sensible for me for that though problems have been discussed from different themes and crosscutting view with clear lists, paper structure still confuses me a little. Many terms like visualization method should be developed for better data mining technologies and new tools & methods should be improved with increasingly high representation technologies are not well explained clearly. Some points of view are overlaid when illustrating those challenges. Why representation, integration of computing and interfacing, Interface and usability are the four major theme and what makes them distinctive and related with each other are not well explained. Challenges referred in this paper about geo visualization are not just limited to the visualization technologies, and these could mostly be challenges faced for many concepts in GIS, discussing about data format problems, data amount problems, AI application issues, human-centered, etc. what are issues should also be discussed and think about for term like spatial statistics analysis methods development. I am wondering how to balance the information accuracy (value) and the interface friendliness. Also, Is the geo visualization always the final steps for data analysis, making results more understandable for further use? Will geo visualization technologies be more important dealing with data and information itself or just focusing on results to be better represented?

In this paper, Koua et al. developed a geovisualization use and usability assessment method based on variables including user tasks, time to complete tasks, usefulness and user reactions, compatibility with the user’s expectations for the different tasks, flexibility, perceived user understanding of the representations used, user satisfaction, and user preference rating. Their result seems to be decently analyzed and explain in a understandable way that different geovisualization methods have its advantage on certain tasks, and disadvantages on some of the other tasks. This study enlightened me that geovisualization process as tools to interpret data, rather than a representation of the data. As well as systematically assess and provide the which geovisualization method is better to use in terms of expected tasks it will perform. This helps me to make the decision when it comes to choosing method of geovisualization, which potentially means what kind of tasks I intend to provided for viewers, and what viewer will expect to gain and utilize the data.

However, I do find their assessment design not that convincing, in terms of participants involved in this assessments are all academics or researchers in science related field. Not only I am not convinced that this assessment are only made for professionals, since they exclude the general public, they also exclude policy makers, urban planners, and social activists, who are also potential users of geovisualization product. And sometimes general public and those social science related professionals tend to use more geovisualization products, due to the lack of programming or statistical skills to process and analyze data. Thus, I would argue the flaw of this assessment process is they fail to include all potential users of geovisualization products, when they only choosing participants from the nature science professionals.

VoPham et al. basically summarized the major trend of practical application of geoAI (mostly machine learning, deep learning, and data mining), and its specific practice regards to environmental epidemiology. By using the example of Lin et al.’s air pollution concentration study at Los Angeles, the authors illustrate how geoAI is used to processing the combination of big data from different sources, as well as efficiency computational process on pattern detect and modelling.

However, a question struggles me from their introduction to geoAI in practical use to the end of their envision of geoAI’s future: what is the exact difference between machine learning/data mining algorithms and geoAI. Is geoAI merely a combination of different machine learning or data mining algorithms? Or is it something more complicated than they illustrated in their article? Since from their example of modelling air pollution, Lin et al. (as the authors of original study) says that specialized geospatial data experts are still needed to decide what kind and quality of data can go in the modelling, to avoid the “garbage in garbage out” situation. To me, however, if a geoAI cannot reach a standard to identify and evaluate what should be included in the computational process, it is just a combination of different computational algorithms. Self-evolution and decision making process might be key to distinguish geoAI and combination of algorithms.

Some may argue that geoAI is only on its early stages, and so much more need to be done in order for geoAI to self-evolve and make decisions. However, if geoAI cannot be adaptive to different spatial or temporal instance, what is the need for an AI instead of a team of machine learning programmers and data miners? I believe reaching proper self-evolutionary ability to adapt different spatial and temporal instances, as well as making decisions of what comes in the modelling process, and what parameter or logic need to be change to adapt the different input variables is essential to call it geoAI, rather than a systematic geospatial data modelling algorithm.

Mennis and Guo’s work generalized the trend, progress, and achievement on Spatial data mining, processing, and interpreting till 2009. It is a very helpful review for those who are not familiar with most of the techniques and approaches in the field of spatial data mining. Their work especially focus on the spatial classification and prediction, spatial association rule mining, spatial clustering, regionalization and point pattern analysis. Although this article makes everyone feels so excited about how the boom of geospatial data and mining technique, which feeds into research field, private sectors, and sometimes government operation, opportunities comes with a cost.

I am not saying more available geospatial data is bad, however, there are certain challenges the authors fails to discuss in detail. First is the selection bias when mining spatial data. For people who aware of GPS tracking devices and do not want to share their geospatial data, and those who have not access to GPS tracking devices yet, they are excluded in some of the hottest geospatial data mining realm, such as social media spatial data mining, there is a selection bias with the data, which may leads to unintended exclusion of population from the interpretation of the data. Although it can be taken care of if data from various sources can be joined together are used in the processing and interpreting stage, it is definitely something spatial data miners should be aware of.

Second is the privacy issue, more geospatial data does not actually makes everyone happier, it has a cost. Although more and more geospatial data is masked to protect privacy, the huge amount of data flows inevitably expose some or most population under privacy crisis. Thus there has to be an awareness for data miners to protect study subjects or data contributors’ privacy, and proper supervision in this field need to be address to prevent malicious mining of geospatial data.

At last, the availability of seems infinite geospatial data is thrilling for people who works in this field for sure. However, it also increases the difficulty and skill requirement for data miners. It is not only computational skills that allows data miners to mining the data. More importantly, is the skill to discover, to observe, and formulating the right question, which until nowadays is still heavily depend on human to make the call. Also, the ability to look at geospatial data critically is necessary. Unless the data fits perfect with our questions, there are uncertainties need to be address rather than blindly trust the data because the size of it.

Dunn & Newton offer a pretty decent summary of the most basic form of optimal/shortest path algorithms, and what they say still holds true (at least to my understanding …) some 27 years later. They describe how Dijkstra’s and the “out-of-kilter” algorithms determine routes, the different uses for each of them in emergency planning, the computational requirements, and finally some potential developments in the topic. Although a verbal description of the contents of a matrix isn’t the most intuitive or interesting to me, I found the easy enough to follow and I’m glad they included the few visuals they did.

It is interesting to see what they list as future development opportunities, as most of these things are now incorporated into every current mapping site, public transit app, and dashboard GPS. Even their “most difficult challenge”, temporal change, has been met, and I can map myself to Saskatoon in a few seconds without having to worry about getting stuck in heavy traffic or road closures. It is honestly very difficult to think of something else that I would want to incorporate into shortest path algorithms or an application like Google Maps. My first two thoughts are pedestrian traffic and incorporating greater local knowledge for more complex decisions. i feel like pedestrian traffic could be done very easily through traffic projection just like street traffic, and I’m sure someone has at least tried to do it. On local knowledge, I guess I mean if you want to ask something like “find me the most scenic drive”, “avoid routes where there are frequent accidents”, “only take me down streets with ample night lighting”. Some of these could lead into problematic discussions surrounding the ideas of aesthetics (who decides?) and safety (for who?).

The authors of this article explains and using experiment to illustrate how the shortest path algorithm (Dijkstra algorithm) and the more complicated “out-of-kilter” algorithm fits in the field of network analysis. Although the article is written in 1992, it still provide the basic knowledge to those who are not familiar with network analysis before. It also reflects the most basic level algorithm for nowadays more advanced network analysis algorithm.

One thing that catches my notice throughout the whole article, is that although the “out-of-kilter” algorithm the article mainly tested focus on shortest path, in realistic application, I would argue the time efficiency are way more important than the distance between two routes, in the case of emergency evacuation. As the authors discusses, however, more factors are indeed needed to perform a network analysis that considers time efficiency, such as peak hours, route capacity, means of transportation, slope, landscape etc..

Another important issue from this 1992 article about network analysis is the limitation of computing power and technological foresight. Due to the computing power limit, only simple shortest path between nodes can be compute in euclidean distance, which in most of the case, will hit obstacle in real-life practice, due to multiple physical, social, and cultural barrier. The issue with technological foresight is that although network analysis itself does not necessarily require geographical coordinate to perform. Or in other words, network analysis itself is not limited by geographic coordinate system or projection. However, nowadays, when we apply network analysis on more advanced use, like GPS tracking, navigation, even real-time traffic monitoring/dispatch, the geographic side of network analysis just cannot be totally ignored, and sometimes rather necessary.