User:Phlox/aphorism: facts or theories or beliefs

From Familypedia

The collision of structured knowledge based technology with the field of genealogy may leave a bit of car wreck. When Joe Friday shows up to ask for "Just the facts Ma'am", he will get some blank from nearly every genealogy enthusiast in the room. The professional genealogist may be one of the few voices that can respond meaningfully to such Joe Friday kind of Guys that have one model of reality. These folks go to great lengths to collects the evidence, and this is a good measure of the quality of a genealogy database. The reality is though that this activity is not all that fun for most people interested in genealogy. Many are very interested in finding out something of their ancestors, and marvel at the breadth of the connections with their fellow kin. We all share a common ancestor who walked the earth just 60,000 years ago.

As we look closer, we first stumble over some skeletons in the closet. A mother whose parentage was purposely altered due to racism, a great great grandfather who had pretensions of being from a manor born whose real father was not of the desired class. Then there is the proverbial "your father is the milkman" problem. After a while, the lineages resemble not some factual structure like a biological genus/phylum hierarchy, but a set of shared beliefs about one's past that in some families is an important narrative, regardless of what the facts are. Professionals want to shatter these myths and get at the truth, and the controversies over genealogies are so contentious that even the Christian bible admonishes^[1] followers not to fall into the trap. We can support both set of users.

• For the set of users that want to get at the facts, we provide information as lists of working theories each supported by empirical evidence with which we have varying levels of confidence (high for dna evidence, low for family tales that a grandmother remembered hearing as a child). Theory A is that one person is the father, and Theory B proposes alternate parentage, with an equally compelling set of evidence. Instead of a single "info" subpage, we have multiple "theory" subpages. The other way is not to draw them together- to just represent them entirely separately as if they were different individuals. That turns the database into a bag of perpetually replicated lineages where each gedcom uploaded posits new individuals that may already have 30 other instances with largely identical information.

• For the set of users less interested in these finer details, the article populates the infobox with whichever theory page gains consensus as being the most plausible. They get a sense of what their lineage most likely is, and later may or may not choose to dive deeper by looking at the alternate theories of certain individuals where there is greater controversy. The door is open, but the front matter of each article can provide and informal introduction about what is commonly believed to be true about the individual. In this way we do not drive users away by too early burdening them with with this complexity.

• Thirdly, some folks don't care about the so called facts or are for whatever reason find the dominant theory on some individuals unacceptable. Some are due to disputes over particular sources that some believe are untrustworthy. Others have to do with a linkage between their sense of identity. For example, some may take deep pride in thinking of themselves as French, but later discover evidence that their ancestors were really immigrants from elsewhere^[2] who found it expedient to explain their odd behavior by saying they were from France. We need to be able to preserve those particular narratives no matter how unlikely.

Representing shared beliefs is considered by some to represent the state of the art in knowledge based systems. Rather than a KBMS approach that assumes a single version of truth that works well in technical subjects, we assume that each assertion has confidence associated with it- confidence that is associated with the weight of evidence, or may be arbitrarily overridden.

The mechanism for overriding layout representations is CSS style sheets. Similarly, we will eventually be able to allow users to present their version of the truth by allowing users to set a preference file that overides the default choices of information pages used to populate certain articles^[3] The collaborated, common view may be different, but those in the minority about which theory should be dominant are not shut out of the collaborative process simply because of those particular differences over one or two articles.

1. ^ Epistle to Titus 3:9 "refrain from investigations of foolish questions, from research into the histories of the genealogies, and from quarreling and disputes about the Law; there is no value in them; they are pointless".
2. ^ For instance, coneheads.
3. ^ Technically, this might be like a belief net overlay, where a set of confidence weights overrides the dominant ones in order to model an alternate scenario. Practically, it may be a lot more brutal in early versions- eg, for article X, Y, and Z, use data from page X/theory_58 subpage, Y/theory32 subpage and so on.