This paper presents two data-driven activities aimed at secondary school students to promote data thinking skills, the first focusing on data literacy and the second on computer programming with a 'data-centric' pedagogy. We briefly introduce a Python package, called ToyPandas, that we developed to make coding more accessible to novice data science enthusiasts. The paper then discusses the motivations and challenges of providing an introductory data activity for students without a computing or data science background, and presents our experience, discussing some preliminary assessment results.
Data science is rapidly gaining popularity in all fields of application, from science to art,
engineering to business, and so on. Even our daily lives depend on decisions based on large
amounts of data, which are seen as a resource for discovering valuable knowledge. This has
created new business models, services, and jobs, and increased the demand for professional data
scientists. As a result, data science curricula have been required in many disciplines in recent
years [
Data literacy can be defined as the ability to understand and critically evaluate information
derived from authentic data [
In this discussion paper, we present two data-driven introductory learning initiatives that we have carried out over the past year to introduce data literacy and promote data thinking to non-major students. In particular, we describe two activities aimed at secondary school students to introduce them to data visualization and data transformation, introducing Python programming as a means to perform simple data science tasks. The remainder of the paper is organized as follows. In Section 2, we present the motivations for ofering introductory data science activities for school students, as well as some challenges in ofering such activities. In Section 3 we present some related work. In Sections 4 and 5 we describe the design and setting of the two data-driven initiatives and present some preliminary results. Finally, we conclude the paper in Section 6.
As mentioned above, as we live in the data age, it is fundamental that people have some basic data skills in order to be active digital citizens. Students should be aware of the challenges of data ethics and learn how to share, handle, communicate, and interpret data appropriately, regardless of their future careers. In addition, we believe that early exposure to data science topics in school can help to embed data culture in students and make them aware that data science has become ubiquitous due to its domain-specific nature. This is also reflected at university, where many non-computing academic programs (e.g. biology, business, medicine, etc.) have started to ofer courses that cover data analytics topics, and students need to be aware of this. Furthermore, it is widely recognized that data thinking, like computational thinking, should be introduced to students as early as possible to spark their interest and create a potential future generation of data scientists, or at least data literate people.
These are the main motivations for ofering data dissemination activities to our students.
If we add the fact that it is dificult to find experienced data science teachers or support
materials for data science education in (Italian) schools, the issue becomes even more important.
Furthermore, data science can be a means to promote and improve competency-based education
with an interdisciplinary approach. The use of real data can be exploited to design learning units
-"Unità di apprendimento" (UdA) in Italian-, creating learning pathways that mix humanities,
sciences, coding, and digital skills. In particular, authentic data can be relevant to students’
interests and attitudes, which can have a positive impact on their creativity and motivation
to learn. In addition, real data sets have been shown to promote data thinking in learning
environments designed to develop computational thinking [
On the other hand, early data science activities lead to a number of challenges that need to be considered by educators and researchers. Firstly, achieving the desired learning outcomes, as data science education often requires some prior knowledge of computing and statistics/mathematics. Coding and statistical models are important components of data science, and students typically need to learn a programming language and/or some statistical packages (e.g., R, Python, Pandas, NumPy, matplotlib, etc.). The domain-specific nature of data science is certainly an opportunity for learning, but could also be a challenge if not addressed in the right way: domain knowledge should be carefully assessed by choosing a dataset that requires a "right" level of knowledge, and with a view to an UdA, teachers from non-computing disciplines should contribute their domain knowledge. Finally, time and curriculum constraints could be another important barrier to the dissemination of data science in schools. In our experience, activities have to be embedded in a very short time frame of 10/20 hours maximum, which leads to several challenges in terms of providing the right balance between learning content and hands-on practical activities. The latter is a major component of data science education, but often requires a lot of time, as non-computing students often struggle with the technical and coding aspects. Contribution. Our contribution aims to address the aforementioned challenges by designing two activities aimed at all secondary school students regardless of their background. We present a Python package based on Pandas, called ToyPandas (working title), to simplify the syntax and the notional machine (NM) behind Pandas, in order to focus only on the most important aspects for an introductory course in data-centric programming. We will also present a data literacy activity based on a Business Intelligence tool (e.g., Tableau) to introduce students to some relevant data topics, such as data analysis, interpretation, and visualization. Our aim is to give students a first experience in data science and, more generally, to make them aware of the importance of data in the development of a better society.
Although data education is still in its infancy and rarely integrated into school curricula [
Our interventions are less ambitious than the above-mentioned curricula/courses, in fact, they are much shorter/compact in terms of learning content, due to the rather tight time constraints. Our initiatives are mainly designed for school Pathway for Soft Skills and Educational Guidance (known in Italy as PCTO). Nevertheless, it is possible to make some reflections/comparisons that will be presented in this paper in the respective sections dedicated to each presented activity.
We designed a teaching format on data literacy aimed at secondary school students that focused on data analysis and visualization. We have experimented with this unit twice, namely in a.y. 2021/2022 and 2022/2023 with grade-11 students of a high school major in humanities ("Liceo Classico" in Italian)8 and was part of a school PCTO activity. In particular, students were enrolled in the European Classical Lyceum curriculum, which, compared to the traditional one, includes the study of economics, law, and two foreign European languages, as well as the strengthening of scientific disciplines.
The students had no previous background in computing or data science, so this was their first data-driven experience (taking into account both formal and informal learning). The teaching unit lasted approximately 20 hours, with 5 hours of lecture and laboratory, approximately 12 hours of a homework team project, and 3 hours of final plenary discussion/presentation with classmates and teachers.
Both editions used Tableau9 software, a BI tool that provides a simple and colorful user interface and enables the creation and sharing of accessible dashboards and visualizations. Although there are many alternative solutions, including educational and open-source options such as Orange10 and Inzights11, we chose to use Tableau, primarily because of its intuitive drag-and-drop interface, which is particularly beneficial for our target audience of students with limited digital skills. In addition, Tableau ofers a wealth of online resources and support materials that further facilitate the learning process. Finally, we did not consider programmingbased solutions such as Python, R, or JavaScript with the D3.js library because the students had no previous experience of computer programming and acquiring such skills in the limited time available would have required a significant cognitive efort.
Unit Overview. The unit consisted of a five-hour lecture (seven for the first edition), consisting of a frontal lesson introducing data and a laboratory lesson using Tableau, a team project that students were expected to spend 12 hours a month on, and a final meeting where student teams could share and discuss their findings. In the lesson, we introduced students to the following topics: Introduction to Data Science (e.g. what data science is, big data as a valuable resource, 8The Italian "Liceo Classico" (literally "Classical Lyceum") is a high school specializing in classical studies, where students study both Latin and Ancient Greek, with advanced curricula in philosophy, literature, and history. 9https://www.tableau.com/it-it/products/public We recommend Tableau Public, a freeware solution that only requires an e-mail account and is available both as a computer programme and as a cloud web-app. 10https://orangedatamining.com 11https://inzight.nz the 5V model, what a dataset is); basic Tableau data types (numerical, textual, geographic) and qualitative/quantitative variables; data ethics (basic concepts of data fairness, security, and privacy); data exploration and visualization (history of data visualization, summary statistics, basic chart/plot techniques). The first edition also covered data cleaning, where students could learn about data quality issues and their causes.
During the hands-on activity, we used a toy dataset and emphasised the importance of efective communication and storytelling techniques when presenting data insights. We also encouraged students to consider ethical issues around data, such as privacy concerns, as Tableau Public shares all results with the wider community.
In the team project, using a collaborative learning approach, students had to identify a research question based on a real dataset12, and from there to create and interpret appropriate visualizations to generate new information. Finally, they had to prepare a multimedia presentation to showcase their own data-driven project and discuss the technical expertise and the interpersonal skills they had acquired. The dataset was selected by us, in particular for the first and second edition of the activity we proposed the Olympics dataset13 and the Economic Freedom Index dataset14 respectively. Both datasets included various geographical variables, giving students the opportunity to explore and experiment with Tableau’s geocoding features such as heat maps (chloropleths). The Olympic dataset required historical-sociological knowledge to explore some aspects (e.g., the evolution of women’s participation or the impact of the World Wars on the competitions), while the Economic Index dataset required some economic/financial knowledge to understand the diferent indicators. This presented an opportunity for students to apply and deepen their understanding of subjects they had previously learned in other disciplines, such as economics, law, and history.
Preliminary Results. At the end of each edition, we collected team projects and responses to a final satisfaction questionnaire. The questionnaire was individual and consisted of the following questions: (DS) How interested were you in the unit? ; (H1) How challenging was it? ; (H2) How much fun was it? ; (H3) How useful was it?. All items were rated on a Likert scale from 0 to 3, with options ranging from "not at all" to "very much" (note that H1 is a reverse item).
In particular, the first edition had a population of 28 students, while the second edition had 42 students. The first session was held remotely due to the Covid-19 health emergency, while the second session was held in person in a computer laboratory at the Department of Computer Science of the University of Genoa. The team projects were evaluated according to the following evaluation criteria: presentation quality and time management (multimedia presentation/slides, adherence with assignment deadline); data analysis (numbers of variables considered; types of data operations such as drill-down, filtering, etc.; the depth and structure 12In the second edition of the course, the research questions were pre-assigned at the request of the school. 13https://www.kaggle.com/datasets/heesoo37/120-years-of-olympic-history-athletes-and-results The Olympics dataset covered all the Games from Athens 1896 to Rio 2016 and contained more than 271 thousand rows and 15 columns. Each row corresponded to an athlete competing in an individual Olympic event and contained basic bio data on athletes and medal results. 14https://www.heritage.org/index/ The Economix Freedom Index dataset measures the economic freedom of 184 countries based on trade freedom, business freedom, investment freedom, and property rights. Specifically, we combined five years of observations from 2018 to 2022 into a single dataset to allow students to analyse the global economic prosperity over the past five years. of the analysis in relation to the original research question); communication with data (efective use of maps/charts, presence of labels/legends, interactivity in dashboards, etc.), and domainknowledge (quality of in-depth study, cited sources, etc.). The scores for each dimension were given in decimals, and the final score was calculated as a weighted sum of the diferent criteria. Presentation, data analysis, and communication had the same weight (0.3), while domain knowledge had a weight of 0.1.
In the first edition, all team projects were of good quality and students seem to appreciate
and grasp the potential of a data-driven approach as well as its application in their personal
and professional daily lives, as already presented in [
Figure 1 shows the results of the final questionnaires. Only 6 and 11 students in 2022 and 2023 respectively completed the questionnaire. The median of the item DS is 1.5 [2 in 2023], while the medians of the three items related to the homework (H1, H2, H3) are 1, 2, and 2 respectively [1, 1 and 2 in 2023]. In general, the activity was perceived as dificult in both editions (all H1 responses were negative), but more so in the second, although it was perceived as more useful. The level of interest in the activity is quite similar, although the first edition had a higher percentage of very positive responses. We did not conduct a statistical control between the two sessions as several variables were changed (remote vs. face-to-face, data cleaning phase, dataset change, pre-set research question), some of which were beyond our control.
In an informal way, we can measure the variability between the two editions based on the
anonymous free comments from the 2023 questionnaire. Specifically, about half of the
participants expressed a preference for working with a diferent dataset (one that is not economic/legal)
and/or having the opportunity to propose their own dataset. Two respondents would also have
liked a lesson on multimedia presentations. Finally, one person suggested splitting the theory
and practical sessions into separate days. We believe that a data-driven activity will spark
curiosity and motivation if the topic is of interest to the student, so in the future, we will
propose a number of datasets to choose from and leave the students free to devise their research
questions.
This second contribution stems from a desire to experiment with data-centric computing
pedagogy in computer programming [
ToyPandas. ToyPandas is a Python package based on Python Pandas, which pedagogical aid
is to make programming more accessible to beginners thanks to a less complex NM and an easier
to learn syntax. The syntax is simplified but similar to that of the original Pandas, so students
can learn the basics and then move on to Pandas, finding a familiar environment and syntax,
without having to create a new mental model of the NM from scratch, but rather enriching
their existing one. For example, to create a dataframe starting from two series:
s1 = series(2, 0, 5, 9)
s2 = series("Bob", "Alice", "Ted", "Carol")
df = concat(s1, s2)
df.show()
The syntax and the NM behind are simpler. In original Pandas, you can define series starting
from iterables, dictiories, or scalar values (e.g., with np.arange()), so the original instruction
might be something like s1 = pd.Series([
Another distinction lies in the data types. In our library, no prior knowledge of Python or NumPy iterable data types is required. Text and heterogeneous series are referred to by diferent names - "string" and "object" - rather than using a single identifier called "object". Integer and lfoating-point columns are simply referred to as "int" and "float" respectively, without specifying the bitsize of the type in the column name (e.g., int32, int64, float64 etc.). In general, the syntax of ToyPandas is more expressive and provides clearer semantics. For example, we have simplified the way we access the dataframe elements (.indexlocate[] instead of .iloc[]). We also made no distinction between .loc[] and .iloc[]. Students can also learn about iteration, functions, and Boolean conditions by applying the function to an entire series or by querying one or more columns of a dataframe (using the bitwise operators) using the original syntax (e.g. df[df[(Age > 18) & (Name == "Alice")]]). The iteration over rows is somehow implicit (a bit like SQL queries)15.Finally, we have overridden several Pandas methods for series and dataframes, including those for handling missing values (NaN).
From a technical point of view, we created subclasses of Pandas classes (e.g., Series, Dataframe, pd.core.indexing._iLocIndexer etc.). These subclasses simplify tasks such as column renaming, duplicate removal, missing values handling, element access and function calls (arguments, return values, etc.). Furthermore, our classes remain compatible with data visualisation using Matplotlib.
Our package is still a work in progress, but we aim to publish it on PyPI by summer 2023. In April we will test it with 30 secondary school students in a PCTO activity. The unit will consist of two lectures (10 hours in total) and will cover the basics of data and computational thinking: Introduction to data science (what is data science, data ethics); Introduction to algorithms and programming (flowchart, program, executor, etc.); Basic data types (int, float, bool, string, series, dataframe) and their main operations; Starting with data (loading data from .CSV files, understanding data summary); Data exploration and analysis (filtering and retrieving relevant data using conditional expressions with logical operators); Functions (applied to columns, to transform data); Data visualization (using Matplotlib). We plan to evaluate the activity in terms of the pedagogical efectiveness of our personalized, data-centric approach.
We presented two data-driven initiatives to promote data literacy and data thinking in secondary schools: the first was based on a data visualization tool, while the second was based on Python programming using a proprietary package to help students perform data science coding techniques more intuitively. We plan to release this package by summer 2023.
As current and future work directions, we mention activities tailored to stress ethical aspects, e.g., in terms of coverage, when applying data transformations and an ongoing project From botany to Big Data16 to develop dashboards analyzing environmental data acquired in various forms observing the plants in school gardens, so to demonstrate how to use digital devices for making the school garden a laboratory for the environment. 15More specifically, many Pandas operations are vectorised to improve performance, but this optimization detail is beyond the scope of our introductory intervention. 16competenzedigitali.unige.it/pon-green