# How to translate "bias"

I am going to hold a presentation in German on maximum entropy (ME) methods and a crucial concept for understanding the principle behind ME is bias.

For example, for an experiment with outcomes A and B without any more information, assuming that they occur with 0.5 probability each is the least biased assumption to make. How should I translate this concept into German?

I thought about Voreingenommenheit, but I would like to double check with somebody who maybe already has experience with the concept.

• As a half baked answer: there is a rising number of Anglizismen in German. So the word "Bias" might be understood by your peer group. Voreingenommenheit is a profound translation. Additionally I would use Paradigma - unlike a Theorie the assumptions are mostly unware of. dwds.de/wb/Bias and dwds.de/wb/Paradigma (thus I do not comply with the definition, more with the examples) Dec 12, 2020 at 12:33
• Possibly helpful: de.wikipedia.org/wiki/Systematische_Abweichung Dec 12, 2020 at 13:34
• In colloquial terms, when first learning the word,, I used to understand beeinträchtigung. Suppose the similar Anlaut and its being used in input-output diagramd influenced (beeintraechtigt) this Dec 12, 2020 at 17:58
• On another note, it should be possible to translate the whole thing to the frequentist unterpretation (since somebody mentions LaPlace in an answer). 50:50 would in that sense be Normal (normalste) and barely even worth a comment. Fair is another word. Ausgewogen (balanced, fairly balanced). Never understood this Lord Bias wank to be honest. Dec 12, 2020 at 18:04

Certainly not Voreingenommenheit in the context of probability theory. When talking about estimators, it is common to call unbiased estimators erwartungstreu, and then for biased estimators there is only nicht erwartungstreu, as far as I know.

Then, there is also Bias as in Bias-Variance-Tradeoff, which remains untranslated (Wikipedia says Verzerrung, but I have never heard that being used in practice).

• Yes, so probably not translating the term at all is the best way. I had also thought of that, but I was unsure. Thank you and @David Vogt for the suggestion. Dec 13, 2020 at 17:27

Technical terms generally do not need to be translated. In fact, many scientists, having grown used to a non-translated term, will perceive the native equivalent as less precise than the original and prefer, for instance, Bias to Verzerrung.

This is of course somewhat of an illusion generated by a differences in exposure. The non-native term is encountered mainly in technical language, whereas the native term is not limited in this fashion.

In this case, the point is moot insofar as the technical term has become a normal German word. It is listened in dictionaries (DWDS, Duden) and capitalised. Therefore, I would suggest using der Bias. Masculine gender seems to be more established than neuter, which is easily verifiable by Googling.

• How is Bias to be pronounced (when used in German)? Dec 13, 2020 at 17:09
• @PeterMortensen Meiner Erfahrung nach [ˈbaɪ̯əs]. Dec 13, 2020 at 18:03
• Technical terms can quickly become terribly illdefined when used out of language, in my experience.
– Karl
Dec 13, 2020 at 18:22
• "This is of course somewhat of an illusion generated by a differences in exposure." - on the other hand, it sometimes is not: The loanword tends to be imported for quite a specific meaning and then adopts that specific meaning (even if it was not that specific in its original language!), while the equivalent native term is, if not less specific as such, at least already in use to express different things depending on context. Case in point, the term "Verzerrung" can easily be mixed up with its meaning related to geometrical distortions (in an image or similar), whereas in German, "Bias" ... Dec 13, 2020 at 23:58
• ... is firmly linked to the idea of data in a scientific evaluation being influenced by factors that were not meant to be a part of the data retrieval or generation. Dec 14, 2020 at 0:00

I'd translate bias into Ungleichgewicht in the context of ME.

Although, in your sentence "... with .5 probability each, is the least biased assumption to make.", maybe you could simply say "most realistic" or "closest to reality" in German, which would translate to: "dass sie mit .5 Wahrscheinlichkeit auftreten die realistischste / zutreffendste Annahme."

Hope that makes sense.

For the non-biased situation, i. e. all results having the same probability, the term

Laplace-Experiment

is quite established. (Alternative: Gleichverteilung). I find it easier to understand, than Bias which has a strength and a direction requiring additional explanation. The German Bias has some disadvantages like possibly unknown (to the auditory) and hard to distinguish inflection forms. Also the English word has other meanings like Befangenheit, which the German variant seems to lack.