Adapting to the test: performing algorithmic adaptivity in Danish schools

ABSTRACT

Algorithmic practices are becoming increasingly more central within educational governance. By focusing on the mechanisms of a particular algorithmic testing system in Denmark, this paper highlights how such practices are implicated in the emergence of new accountability infrastructures. It adopts an STS approach drawing specifically upon Michel Callon’s concepts framing, overflowing, and re-framing. The paper examines how algorithmic adaptivity has become central in the framing of the Danish national test and traces the ways in which students, teachers, and schools respond to such proceduralized interactions. While algorithmic adaptivity was introduced as a way of providing students with an equal test experience, it also inscribes student adaptability into test practices, generating new student affectivities and teacher responsibilities in the process. The paper argues that this is a matter of adapting to the test and highlights how the mundane practices of testing situations also become a subject of governance.

KEYWORDS:

• Computer adaptive testing (CAT)
• digital educational governance
• performativity
• algorithms
• STS

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. I will not be using the abbreviation CAT in this paper, but instead refer to ‘the adaptive test’ and the algorithm. This allows me to draw more specific attention to the adaptivity of the test.

2. Throughout the paper, I will use the terms performed, enacted, and done interchangeably.

3. Although, I engage with literature from both ANT and what is known as post-ANT in this paper, I will not elaborate the distinction here, but use the abbreviation ANT as shorthand.

4. The Rasch model’s is used in many tests, e.g. international tests, such as PISA, TIMS, PIRLS. For more on this see (Allerup & Sorber, 1977; Allerup & Kjeldsen, 2017).

5. The Danish term ‘Elevdygtighed’ does not easily translate. However, following Allerup, I will use the term ’student achievement’.

6. In addition to the Rasch model, the adaptive test incorporates a very large number of questions or items of different difficulty, referred to as an ‘item bank’.

7. Most of these tests, however, were not used for educational purposes. See (Allerup & Kjeldsen, 2017) for a list of specific publications concerning adaptive testing.

8. This includes the way in which the test chooses the first items of the test, each subsequent item, and the criteria used in deciding to stop the test.

9. It is possible to adjust these settings to vary the ratio. A 20–80%, split, used in other CATs, would, for example, give the students a better chance of solving the items (Allerup & Kjeldsen, 2017).

10. The Danish national test provides test scores on two different scales: a criterion-based scale and a norm-based scale. The calculations that go into producing the scores are not the focus of this paper. For a discussion of test scores see Ratner (forthcoming) Europeanizing Danish Test Data: Enacting students, populations and futures.