torsdag 14 maj 2020

Skill rebound (paper)

This is the second paper of mine that was accepted to LIMITS 2020

My previous blog post was about a paper of ours that has been accepted to the upcoming Sixth Workshop on Computing within Limits. As it is, I submitted a second paper that has also been accepted, "Skill rebound: On an unintended effect of digitalization". The paper was written together with Vlad Coroamă from ETH Zürich (he is the first author).

This paper has an interesting history that started with me listening to Vlad's presentation of his paper "Digital Rebound – Why Digitalization Will Not Redeem Us Our Environmental Sins" at the 2019 ICT4S conference. That paper is written by Vlad Coroamă together with Friedemann Mattern and is available as a pdf here).

Vlad's presentation set of discussions between us at the conference and we almost immediately felt we were on to something and started to discuss the possibility of possibly writing a paper together (based on ideas we had bounced between us in these discussions). Looking at the history of the shared Google documents we have worked with, I can see we started to take notes already then and there (at the conference) and that we had a follow-up meeting at the end of Augusti. Then there was a hiatus for half a year when nothing much happened and we only stated to work on the paper in earnest in February.

This paper could easily have been aimed for and submitted to this year's ICT4S conference, but we were both busy at the start of the year and did not have any chance of meeting the early February deadline, so we instead aimed it for Computing within Limits.

The paper's backstory is that we had the idea that while self-driving cars/autonomous vehicles (AVs) will make each trip more energy-efficient, AVs will surely also lead to more trips being made. If driving becomes safer, easier, more convenient and less expensive, it's hard to imagine that the demand for (autonomous) car trips will not increase. It's in fact quite probable rather than just, say, "possible". The specific insight we had was that driving an ordinary (non-autonomous) car is limited to those who have a drivers' license, but with AVs, "anyone" can drive (or "drive"). A task that previously required training and a specific skills set will now be open to anyone (including children, pets and inanimate objects such as, say, a violin). If we think about this in terms of "rebound effects", this is a type of rebound effect that we haven't seen anyone else talk about before.

There are different kinds of rebound effects (we go through some of them in our paper), for example "time rebound". We suggest that we describe a different type of rebound effect and have chosen to call this "skill rebound". In this case the "skill bar" that is required to perform an action is lowered and this leads to more of that activity. If the activity in questions (e.g. driving a car) is detrimental to the environment then skill rebound will have "unintended" (but not necessarily unpredictable) environmental effects. For more on this, do read the paper or just start by reading the abstract:

Efficiency gains in economic processes often do not deliver the projected overall savings. Irrespective of whether the increase in efficiency saves energy, resources, time or transaction costs, there are various mechanisms that spur additional consumption as a consequence. These mechanisms are generically called rebound effects, and they are problematic from a sustainability perspective as they decrease or outweigh the environmental benefits of efficiency gains. Since one of the overarching purposes of digitalization is to increase efficiency, rebound effects are bound to occur frequently in its wake. Rebound effects of digitalization have been ignored until recently, but they have been increasingly studied lately. One particular mechanism of digital rebound, however, has been largely disregarded so far: the digitalization-induced lowered skill requirements needed to perform a specific activity. As with other types of rebound effects, this leads to an increase in the activity in question. In this paper, we propose the term skill rebound to denote this mechanism. We use the example of self-driving cars to show how digitalization can lower the skill bar for operating a vehicle, and how this opens ‘driving’ a car to entirely new socio-demographic categories such as elderly, children or even pets, leading to increased use of the (transportation) service in question and thus to rebound effects. We finally argue that skill rebound must be better understood and taken into account in the design of new technologies.

söndag 10 maj 2020

From Moore’s Law to the Carbon Law (paper)

This is the second paper that comes out of the FLIGHT research project

Our second paper from the FLIGHT research project, "From Moore’s Law to the Carbon Law", has been accepted for publication at the upcoming Sixth Workshop on Computing within Limits. All five authors work in the FLIGHT research project (or, formally, "Decreased CO2-emissions in flight-intensive organisations: from data to practice") and they are Daniel Pargman, Aksel Biørn-Hansen, Elina Eriksson, Jarmo Laaksolahti and Markus Robèrt.

I wrote a blog post about the first paper from the project, "On the necessity of flying and of not flying", only one and half month ago and while these two papers have not exactly been written in parallell, they partly overlap (with me as the first author of this paper and my colleague Elina as the first author of the previous paper).

The paper submission deadline was originally at the end of March, but it was postponed two weeks due to Covid-19. We just found out the paper has been accepted and there's now a very short turn-around to improve the paper (taking the feedback of the reviewers into account). The paper itself will be published and made available online sometime later this summer.

The paper was originally meant to be a more narrow description of the problem (researches fly too much) as well as of our project and our proposed solution (e.g. how we work to decrease CO2 emissions from flying at KTH). As we brainstormed and developed the core argument of the paper (before we actually started to write it), we realized we could reach higher and argue for something larger and more fundamental than what we originally imagined, linking the paper to the conference and to what others have previously attempted to do within the area of Computing within Limits.

A key phrase and a starting point for the paper is that a goal of the Limits community is “to impact society through the design and development of computing systems in the abundant present for use in a future of limits and/or scarcity”. The paper's argument in a nutshell is that it has proved difficult to design and develop such systems since we only have very general (and hazy) ideas about what "a future  of limits and/or scarcity" will look like. To develop systems for an unknown future is obviously hard and the alternative is that different researchers have (possibly very) different ideas about what future(s) we are aiming or heading for. It has in other words been hard to find common ground. This paper addresses that conundrum by proposing a roadmap for going from here (the present) to there (the future) in the hope that it might provide a foundation for future Limits papers. The paper was easy to write but we expect it to generate  a lot of discussions (and possibly controversy). Here is the paper abstract:

In society in general and within computing in particular, there has, and continues to be, a focus on faster, cheaper, better etc. Such perspectives clash with the fact that impeding climate change and the need for radically decreased CO2 emissions (c.f. the Paris Agreement) will have fundamental and far-reaching ramification for computing and for all other sectors of society during the coming decades.

In the call for the first Computing within Limits workshop, it was stated that “A goal of this community is to impact society through the design and development of computing systems in the abundant present for use in a future of limits and/or scarcity.” There have since been several contributions to Computing within Limits that have accepted the challenge of discussing and imagining what such systems as well as what “a future of limits and/or scarcity” could look like. Despite this, there is currently no consensus about what exactly such a future entails and the community can consequently only offer hazy ideas about exactly what systems we should strive to design and develop. The basic problem can be summed up as follows: we know that fundamental changes are necessary and will come, but we still struggle with envisioning what a post-growth/decarbonising society looks like and what computing systems need to be designed and developed for use in such futures, or, to support that transition.

In this paper we argue that the work of imagining an actionable “future of limits” could benefit from using the “carbon law” as a starting point. The carbon law is based on work in the environmental sciences and we exemplify how it can be used to generate requirements that can guide the development of computing systems for a future of limits. While these lessons are general, we exemplify by describing a research project that aims to support the KTH Royal Institute of Technology’s goal of - in line with the carbon law - radically reducing CO2 emissions from academic flying over the next decade. We give examples of how computing can aid in this task, including by presenting visualisation tools that we have developed to support the KTH carbon abatement goals. We also discuss the role of computer science in general and of Computing within Limits in particular in supporting the transition to a more sustainable (or at least a less unsustainable) future.