Representativeness of events in timelines
This is a meta page about timelines. View all meta pages about timelines
This page explains the representativeness of events in timelines on this wiki. Timelines can be seen as a collection of events about some phenomenon. It is then possible to ask whether the events collected in the timeline are a representative sample of all relevant events. Unless otherwise noted, none of the timelines on Timelines Wiki are guaranteed to have representative events. As such, questions such as "Has an organization done more things between 2010–2013 or between 2014–2017?" or "Does this movement have more publications by male or female authors?" cannot usually be answered accurately by reading a timeline. Of course, a human with the aid of a timeline is probably still better than a human alone with its mind prone to relying on an availability heuristic.
One case where representativeness is guaranteed is if a timeline exhaustively lists all instances of relevant events. For instance, as far as the creator of Timeline of Carl Shulman publications is concerned, the timeline lists all nontrivial publications by Carl Shulman (as of when the timeline was first written -- Carl has since published more work and these may not all be included in the timeline). Therefore if one were to ask "Did Carl Shulman publish more during 2006–2010 or during 2011–2015?" or "Does Carl Shulman publish more in April or in December?" or "Has Carl Shulman published more on LessWrong or his own blog?", the timeline would be able to answer those questions, subject to those questions being meaningful to ask in the first place (e.g. Carl Shulman might post on a schedule or might complete the bulk of the work on a post in a different month from its publication, so that asking about posting frequency on a monthly granularity might not be meaningful).
Contents
How timelines are constructed
In most cases, timelines are written organically as the timeline writer reads relevant sources, adds events with dates on them, follows references, performs relevant search queries, and so on. This means events that are easy to locate and have definite dates tend to appear more in the timeline.
In some cases, such as Timeline of Carl Shulman publications, where the scope of the timeline is definite and relatively small, all events of interest can be added (at least when the timeline is first being written -- updating the timeline to include new events requires ongoing maintenance). In these cases, an exhaustive search and be conducted, rather than an organic "read some, follow references, repeat" procedure.
Tests
Currently timeline writers do not implement formal tests of representativeness of events. Writers might employ heuristics like "add all events I can think of", "ask experts to see if I am missing something important", "add all grants worth at least $50 million", and "look for similar collections of events and combine them to produce a more comprehensive collection". However there is no formal procedure for such heuristics.
Inter-timeline representativeness
This page has claimed no guarantee of representativeness of events within single timelines, but another kind of representativeness is that across multiple timelines. For instance, seeing both Timeline of libertarianism in the United States and Timeline of the environmentalist movement on this wiki, one might be tempted to ask "Are there more US libertarian organizations or more environmentalist organizations?" Unfortunately, such a query is fraught with even more problems because:
- Two timelines can be written by different people, who can have different standards for inclusion of events. Indeed, Timeline of libertarianism in the United States was created by Issa while Timeline of the environmentalist movement was created by Sebastian.
- Different movements or organizations can have different standards for e.g. publications, making a naïve count of things deceptive.
- When an entity has fewer instances of a thing (e.g. an organization has fewer publications), it is easier to exhaustively include all instances of that thing, whereas if an entity has a great many instances of a thing, it is easiest to include the few instances that can be shown to be significant. This leads to a paradoxical situation where the entity with greater instances of a thing has fewer instances of it included in a timeline than an entity with fewer instance of the thing. An example is Timeline of Niskanen Center versus Timeline of Ludwig von Mises Institute. The former has published a handful of policy reports and articles, so all major publications were added to the timeline. On the other hand, the latter has published hundreds of books and thousands of articles with no clear indication of importance among them, so the timeline skipped adding publications altogether for the first round of work.
- Two timelines can be at different stages of completion. The extreme case is when a timeline of one topic has been written while the timeline for another has not been written at all (a "null string" timeline).
See also
- Inclusion criteria for full timeline in timelines
- Domain names in timelines
- Detail construction for full timeline in timelines
