12 ppl attending
Open exchange
conference
some of us participated in Jena Big Historical Data conference (https://bhdc.earth/): it was good, main topic was the analysis of big data
ONLINE 11th-12th December: Barcelona workshop: Social modelling and simulations: https://heuristica.participa.cloud/conferences/SMSBarcelona2023
free
open online debate space for until 10th December, log in and give your opinion
ArchaeFOSS takes placec next week: https://www.archeofoss.org/2023/
- panel by Joe Roe and Zack Baptist on Social Media and Fediverse in archaeology
various
new: topic of dark data: unused data in databases, usually 60% in a database
the archaeoanalyst by Shawn Graham: https://archaeo.social/@electricarchaeo@scholar.social/111410944058988315 and https://electricarchaeology.ca/2023/11/07/using-gpt4-turbo-to-extract-relationships-and-entities-with-the-llm-python-package/
- Large Language Model to analyse archaeological data (stratigraphy to Harris Matrix)
archaeo-riddle https://theia.arch.cam.ac.uk/archaeoriddle/
talk by Alfredo, Simon and Ismael
Enrico R. Crema und Xavier Rubico-Castillo conceptualised the project
simulate archaeological data for analysis of others -> test archaeological methods and hypotheses, which is usually not possible in archaeology
6 proposals of ppl trying to answer the questions at the EAA conference
modelling process
- world “rabbithole”
- elevation map generated through Perlin noise
- sites of two different groups (farmers and hunter-gatherers) generated on parameters like nearness to water
- archaeological record generated by population dynamics models
- loss of the record (short term and long-term)
a lot of sites disappeared, lost evidence etc –> record of each site is generated independently
behavioural model:
meeting between two groups simulated on nearness and population size -> leapfrog or fight
population growth was modeled (fighting makes the growth values a bit smaller than “usual” growth)
dispersal speed via sea crossing faster than via land
game idea
extract coordinates from sites, and 30 radiocarbon dates of the site + affiliation of site
data availability
- map gridded in 100 grids
- participants got info in CSV files of all the same 5 grids of the map, and then chose 5 further to get more information
- already different ways researchers chose the further 5 grids – sampling method!
researchers were given 3 main questions: relationship between the groups, population tajectories and rates of dispersal?
people analysed their data and prepared answers for EAA 2023
plans
- will be sharing their data and describe the code via Bookdown
- there will be a poll on methodology and participation
- collective thoughts on how to improve archaeological methods
- collective publications
Questions/ Discussion
models and all code is in R, because most analyses are done in R (14C dates etc) –> stays consistent
Agent Based Modelling is possible in R
14C dates were taken as “real world” 7500 calBC data
parameters for loss of data etc were eyeballed, but reasonably, so that the outcome looked more or less realistic
in some cases it was well thought out on population movement of rather younger people, dietary ideas etc
many models were created –> “explorative method”
data given to participants: no archaeological types, but affiliations to hunter-gatherers and farmers?
- would have been to complicated to include types at this point
comment: arcmatnet (https://peerj.com/articles/cs-1419/) might be a sensible extension (model for style / transfer)
what happens if you don’t have 10/100 squares, but if you have all the data?
- there is now the possibility to re-sample the data and run it through the 5 methods people have used and check what would change based on sampling
- it would be interesting to check the generation bias and to compare analysis of the wholeness of data before loss and all the data after loss
how accurate were the colleagues in their results?
- it was just a 15 min of results presentation at EAA, so not too much information yet
- very impressive: all got the rough strokes “correctly”, like the war between the two groups
- not the aim to put people in competition, but analyse the approaches, needs to be respectful
- still needs work to put all this together
- every participant used the methods they knew - so this is interesting, that we don’t chose methods regarding the question but what we know already
- one researcher used a non-mathematical approach, someone used a ABM, spatial distribution modelling, … a variety of methods used
data set is useful for teaching
- has already been used by members of SIG SSLA
model will be re-worked and improved
- is the world re-createable? seed set in the model was forgotten, but all the parameters are fixed (will be remembered in the next version)
- could be used as a bench mark dataset for different methods
comment on the project: a great approach and a groundbreaking idea for checking our disciplines methods
