9-weeks Internship Learning Experience

by Aaron Leung, Princeton University

Posted on August 27, 2020 at 4:50 PM

Week 1

My first week with Nautilus started with an introductory virtual meeting with the supervisors. I was introduced to the organisation, the organisation’s values, how the summer work is structured, and possible projects. Nautilus is an educational technology organisation that aims to develop STEM literacy and educational equity through accessible, personalised learning and innovative policy. Nautilus believes in the value of online education as a way of increasing visibility and accessibility of knowledge to students. Nautilus has developed software that enables students to learn through chatbots and educational games.

The workday is fairly unstructured by virtue of a project-based approach. It is interesting to see how college better reflects the work culture at Nautilus than high school; there is a lot more freedom to pursue my interests and the things I care about. Nautilus offered me a few projects on which I could work. One project would be to analyse educational data from Nautilus’s software to form data-driven insights on how students learn best, which could have implications for research and policy in online education and remote learning. Another project would be to explore the notion of gamification – learning through educational games. Games lend themselves well to motivation, so the application of game concepts to education could be beneficial. Another project would be to develop educational tools and games using Nautilus’s software infrastructure, helping to facilitate and improve online learning for student users. In recent times, the development of online learning and the democratisation of knowledge has vastly improved educational outcomes for underserved populations, and I seek to be a part of the movement – pushing the agenda for educational accessibility and equity.

Throughout the week, I have been familiarising myself with Nautilus’s existent infrastructure. I have brainstormed my role, gauging my skills and interests in the presented projects. I’ve begun preliminary research into topics like data mining, online learning, and STEM gamification. I’ve become better acquainted with learning methods such as Eric Mazur’s learning cycle. In light of my focus on service and education, I would like to pursue a project that has implications for the community at large.

Week 2

My second week with Nautilus started once again with a virtual meeting. I learnt about the value of multiple choice questions in testing and surveys. These types of questions are especially compatible with data analysis, as numbers and defined choices are more easily aggregated and comprehended by programs than unstructured answers. I learnt that it is important to collect data in a way that enables sufficient analysis thenceforth. To that end, I was given access to “CSV” (comma-separated values) files containing educational data from Nautilus’s software and apps. To see what sort of software Nautilus is creating, I took a peek at a few gamified puzzles that Nautilus developed using JavaScript and AI principles. I also explored a few research conferences related to education and STEM – in the event that I turn my findings into a paper.

I am currently in the ideation and preparation phases. My tentative plan is to broadly examine the effects of gamification and online contexts on youth learning and development using sources like educational data and psychology literature, taking into consideration factors like educational equity and the variance in learning styles.

In order to conduct research, it is important to acquaint oneself with the tools and methods typically employed in a select discipline. Initially, the concept of data analysis seemed quite intimidating to me, as I envisioned the integration of concepts like machine learning and complex modelling. However, looking through the CSV files, I realised that they are more straightforward than I expected. Nonetheless, I was expecting a fairly steep learning curve from the start. In continuation from last week, I continued to read articles and literature on online learning and gamification, grasping an overview of the disciplines and exploring areas for novel research. I experimented with software such as programming environments and collaborative notebooks. I learnt the basics of JavaScript and Python programming, which are languages used for Nautilus’s software and data analysis respectively. I downloaded the CSV files from Kaggle (a data science platform). Using Python, I began my search into the vast world of data.

Week 3

My third week with Nautilus started with a virtual meeting. This week, I learnt about how to adapt JavaScript programming to the organisation’s infrastructure, understanding its state machine flow and specified external libraries. For instance, the start and state fields must take in a prescribed number of three and four parameters respectively.

Using Python, I was able to uncover a few key insights from the CSV files. For example, 50.83% of students prefer the largest group size option of 4 people. As well, 66.67% of students prefer the flipped classroom model over standard lecturing. Using these data points, I have developed the insights that students enjoy peer learning and prefer priming material outside of class and discussing material in class with the professor and other students. These ideas developed my intuitions on how students learn. However, when handling data, we ought to note and account for influencing factors and biases like sample size and demographics. For instance, it could certainly be the case that some students prefer learning on their own – a case that is naturally entailed by the variance in learning styles.

I believe the root cause of unequal educational outcomes is the comfort of uniformity. Throughout history, social structures have remained intact by virtue of a principled and efficient approach to decision-making, often leading to uniform thought and unified motivation. Our propensity for stability inhibits change, inciting many past and modern social issues. I have explored this issue on campus through my study of philosophy. Since we are a product of our upbringing, we find it difficult to disassociate ourselves from the values and traditions with which we have been raised. More specifically, teachers and professors often teach based on methods with which they themselves have been taught, establishing a potentially suboptimal tradition. However, these principles of uniformity are hardly transferable to the field of education.

Due to the diversity in factors like background and experience, students acquire knowledge in vastly different ways. One scholar I read distinguishes between four styles of learning: concrete active, concrete reflective, abstract active, and abstract reflective. The scholar claims that although the majority of students prefer concrete active teaching, many teachers use an abstract reflective approach, creating a divide between students not because of their ability to learn but by the way in which they learn. Another common framework is the categorisation of students as predominantly visual, auditory, or kinesthetic learners. Suffice to say, everyone learns differently. Developing individualised learning systems could be key to solving problems related to educational equity. My research is currently focussed on how the flipped classroom model and the integration of technology facilitate the pursuit of individualised learning and improve educational outcomes. In the school year, I had to use LaTeX to format my lab report, so I have accessed this software once again to format my findings into a research paper.

Week 4

My fourth week with Nautilus started with a virtual meeting. This week, I learnt more about the organisation’s involvement with high schoolers. The organisation is creating a STEM-oriented curriculum for high schoolers, aiming to teach both tangible knowledge and instill lifelong values such as perseverance and an inventor-contrarian mindset.

Continuing from last week’s research, in line with the preferred method of learning for the majority of students, scholars recommend active and experiential learning that engage the senses such as small group discussions, in-class debates, and the case method approach. We see here that a lot of these learning techniques require student collaboration, and the flipped classroom model facilitates in-class student interaction by deliberately allotting time in class for communal knowledge assimilation.

There are two paradigms in science education. The objectivist paradigm asserts that knowledge is discovered, while the constructivist paradigm asserts that knowledge is actively constructed. Scholars have claimed that teaching with the constructivist paradigm in mind is more effective because this paradigm accounts for variance in prior knowledge and learning styles. A case for re-envisioning modern learning is the prevalence of the objectivist paradigm in traditional learning and the need to transition to the constructivist paradigm in order to provide equal access to education regardless of background.

Another case for re-envisioning modern learning is the misalignment currently present in the learning cycle framework, a three-part cycle that consists of exploration, concept invention, and application. Exploration consists of tangible data acquisition, concept invention consists of abstracting concepts from this data, and application consists of applying these concepts to new situations. The misalignment exists in traditional teaching, wherein the teacher aims to introduce the concept before priming students with intuitive examples.

In light of human behaviour, a learning paradigm shift is bound to encounter resistance along the way – for institutions, teachers, and students themselves. Regarding the integration of technology, there are interesting connections between these three parties. A study shows that technology-related training on an ad hoc basis fosters positive attitudes from teachers and teacher attitudes towards technology influence student perceptions. Therefore, it is evident that the institution’s targeted adoption and embrace of technology is necessary to generate positive reactions from teachers and students.

I think it’s interesting to think about one’s research from a more personal and anecdotal perspective. I have personally found that using real-life examples (like NemoBot games) before delving into the theory improves my understanding of the material by allowing me to first build an intuitive comprehension of a subject. Doing formal research into learning has granted more concrete concepts and terms to my personal observations, and I could potentially utilise this research to enhance my own learning strategies.

Week 5

My fifth week with Nautilus started with a virtual meeting. This week, I learnt about how Nautilus intends to meld the work of its interns into cohesively themed projects in order to incorporate this work into teaching high schoolers in the coming weeks. At the start of the internship, Nautilus sorted interns into two overarching tracks based on interests and experience: software and data and policy. I understand now why such a structure is useful and effective – high schoolers could learn from and experiment with software projects programmed by the software interns, and the data and policy interns could collect poll feedback and usage data to inform future software development and policy recommendations. I discern a characteristically cyclical pattern – the intrinsic interplay between product development and user feedback.

In addition to exploring the technology that enhances learning, Nautilus also explores the methods by which technology does so. One such method is gamification – the idea of implementing games and their principles into another field (ex. education). This week, I pivoted my research from theoretical learning frameworks to gamification. At the most fundamental level, a game has rules, goals, and rewards. The way in which games are structured enables a user to obtain satisfaction and/or a reward from successfully achieving a goal within a set of prescribed rules or constraints. There are biological motivations behind the user’s satisfaction. In an instance of positive feedback, whenever one completes a task, there is a chemical release of dopamine in the brain, which encourages a renewed pursuit of the task’s completion, leading to the task’s completion, leading to a dopamine release, and so on. Here, we see the inception of a self-perpetuating incentive. The incentives created by games are powerful; if misused, games could lead to addiction and the reinforcement of negative habits. However, consider as well the situation in which games are used for good – creating a healthy mindset and positive habits. It is clear that game creators bear significant responsibility, as their intentions and decisions undoubtedly affect the user. For example, in some games, success is achieved through collaboration, while in other games, success is achieved through competition. Further, games could encourage excessive repetition through the constant flow of rewards, whereas other games could dissipate rewards as the pursuit of a certain goal becomes less beneficial to the user.

Week 6

My sixth week with Nautilus started with a virtual meeting. This week, I learnt about how Nautilus intends to use Jupyter Notebook to interact with high schoolers. Jupyter Notebook is a collaborative space for Nautilus-generated activities, student work, and Nautilus’s feedback. Jupyter Notebook is particularly effective for collaboration and learning related to programming, since users are able to run code within the platform. The platform facilitates the implementation of feedback systems and peer learning strategies. I’ve found that one of the joys of the classroom is the opportunity to learn from your peers, as peers are often able to communicate concepts in ways with which traditional teachers are unfamiliar. Introducing peer learning at a young age fosters a healthy mindset of teamwork and collaboration.

In continuation from last week’s forage into gamification, I sought to identify applications of gamification to learning. Gamification could be implemented from both a lower-level and higher-level perspective. From a higher-level perspective, learning platforms could have rules, goals, and rewards that motivate students. For instance, one could conceive of an incentive structure that prioritises collaboration over competition by having students receive a communal reward for completing a prescribed amount of tasks as a group. From a lower-level perspective, knowing that the majority of students learn best using a concrete active approach, games could introduce concepts in the form of concrete active examples. In fact, Nautilus has already implemented games (ex. Tic Tac Toe) into NemoBot so that students are introduced to STEM concepts through fun and intuitive means.

I think a common fallacy that deters learning is the student notion that perception of a subject entails its reality. For example, if students’ previous encounters with a subject left them with a bad taste, they may attribute their distaste to a lack of passion for the subject itself rather than an instance of uninspiring teaching. Gamification instills learning with a sense of fun and play, allowing students to enjoy a subject even if they don’t have the often cited yet elusive passion for the subject.

Week 7

My seventh week with Nautilus started with a virtual meeting. Moving more concretely, I explored technologies that could work in agreement with my aforementioned research as well as educational equity.

Chatbots are “conversational agents that allow the user access to information and services through natural language dialogue, including text and voice”. They may enhance individualised learning by using AI principles to accommodate different learning curves – facilitated by the chatbot’s capacity to collect and utilise data. They limit synchronous learning by facilitating knowledge acquisition without the need for human intervention. By providing instantly accessible information, chatbots eliminate the dead time elapsed in a traditional asynchronous Q & A system. Using a conversational tone and natural language dialogue, they impart a human element to an otherwise faceless experience. Hobert and Berens’s long-term field study on university students revealed a positive correlation between the programmed inclusion of chatbot “small talk” and student receptiveness to a chatbot-based learning system. In other words, humanising technology may strengthen long-term human-computer relationships.

Personal response systems are instructional tools that enable students to almost instantaneously transmit responses to the teacher via signals. They facilitate student engagement by circumventing the waiting time associated with traditional response solicitation and retrieval. In doing so, personal response systems provide students and teachers alike with immediate feedback on individual and class performance, motivating dynamic teacher-driven adjustments to in-class content and decision making and student-driven adjustments to review material and personal learning methods.

Nemo Bot is a multifaceted software platform developed by Nautilus that utilises AI technology to improve peer learning, STEM literacy, and educational equity. The platform enables chatbot functions, personal response systems, access to an open hub of educational applications, and data collection. Nemo Bot has the ability to serve the roles of both chatbot and personal response system.

Teachers traditionally reward academic performance over improvement. However, educational disparities are not simply an outcome of ability nor talent but also of socio-economic backgrounds and other uncontrollable circumstances. According to the Economic Policy Institute, a student’s social class is perhaps “the single most significant predictor… [of] educational success”. The Organisation for Economic Co-operation and Development similarly associates socio-economic background with academic performance, indicating the risk to educational equity of accelerating student progress based on favourable past academic achievement. The OECD recommends “monitoring information on attendance, performance and involvement” to promptly identify and concretely address educational disparities.

Technology has the ability to mitigate these educational disparities. As a self-improving and optimising technology, AI technology could adapt learning platforms in a way that maximises aggregate student performance – intelligently individualising the learning pace for each student to accommodate different learning curves. Similarly, chatbot databases and personal response systems could inform the teacher of academic concerns to enable swift and student-specific action.

As the internship nears its closure, I have begun to take comfort in ambiguity. At the start of the internship, I was a little worried about the ambiguous nature of the internship, but I eventually realised that ambiguity comes with its virtues – namely, the ability to fully immerse yourself in whatever you’re passionate about. The internship’s work environment also simulates the professional world, in which your work ethic is driven by self-motivation and personal accountability.

Week 8 & Week 9

My eight and ninth weeks with Nautilus started with a virtual meeting. In the final stretch of the internship, I crafted a research paper to tie everything together titled “Resolving K-16 Educational Inefficiencies and Inequities through Novel Remote Pedagogical Strategy”. Of note, that title underwent more revisions than I would care to admit. It was a very satisfying feeling when a lot of the pieces of my research floating around the past few weeks started falling right into place. I decided to format my paper using Google Docs instead of LaTeX, and I based my format off one of the papers that I referenced. I organised my research into the usual sections like Results and Discussion.

To give the gist of my paper, here is an abstract as follows.
“In the Introduction section, the paper describes the author’s motivations in crafting the strategy. In the Background section, the paper delves into background research around learning frameworks, data analysis, gamification, technologies, and educational equity. In the Methodology section, the paper develops a set of strategy parameters in agreement with the background research to ensure a relevant, modern, and sound strategy. In the Results section, the paper presents a novel remote pedagogical strategy that aims to resolve K-16 educational inefficiencies and inequities through innovative principles and technology. In the Discussion and Conclusion section, the paper substantiates the strategy’s elements and offers implementation recommendations to Nautilus.”

To give a snippet of my strategy, here is an excerpt as follows.
“The self-originated QDI (Question, Discussion, Instruction) model systematises the internally flexible synchronous learning component of the flipped classroom model (i.e. class).

In the Question phase, the teacher poses a question (or answerable prompt) to the class. The contents of the phase are informed by a variety of factors such as the teacher’s intuitions and aggregate student performance on the pre-class quizzes and in the Discussion phase – hence motivating the QDI model’s abstract flavour.

In the Discussion phase, students discuss and answer the question (or answerable prompt) in small groups. The teacher may divide students randomly or in light of each student’s individual performance. For instance, the teacher could intermix students on slower learning curves with students on faster learning curves to maximise peer learning. The teacher offers assistance on an ad hoc basis. The teacher collects answers through a personal response system such as a simulated clicker application. If students demonstrate unsatisfactory comprehension, the class enters the Instruction phase, else the class returns to the Question phase.

In the Instruction phase, the teacher instructs on a selection of topics for which the teacher believes further treatment is beneficial. Since the flipped classroom model assumes pre-class knowledge acquisition, the class need not address all of the concepts assigned since the previous class. The teacher may elect to commence class with this phase as an alternative to commencing class with the Question phase – for instance, on the basis of concern for unsatisfactory aggregate student performance on the pre-class quizzes and/or previous work.”

This summer, I was able to pursue so many passions of mine: education, technology, and service. Incorporating all three passions, I crafted a formal research paper that synthesizes research and data to create actionable educational policy. Spending a summer immersed in research has enabled me to develop a broader conceptual understanding of service; even without direct interaction, my work could still really positively impact my target community. I am very grateful for the growth and opportunities that the internship has afforded me, and it has been a privilege to work with such a fantastic team over the past few months. Thank you to Nautilus for an absolutely wonderful Summer 2020!


  1. Christensen, Rhonda. "Effects of Technology Integration Education on the Attitudes of Teachers and Students." Journal of Research on Technology in Education 34, no. 4 (2002): 411-433. doi:10.1080/15391523.2002.10782359.
  2. Følstad, Asbjørn, Theo Araujo, Symeon Papadopoulos, Effie Law, Ole-Christoffer Granmo, Ewa Luger, and Petter Brandtzaeg, eds. "Chatbot Research and Design." Springer Nature Switzerland AG, 2020, 1-272. Accessed July 28, 2020. doi:10.1007/978-3-030-39540-7.
  3. García, Emma, and Elaine Weiss. "Education Inequalities at the School Starting Gate: Gaps, Trends, and Strategies to Address Them." Economic Policy Institute. September 27, 2017. Accessed July 29, 2020. https://www.epi.org/publication/education-inequalities-at-the-school-starting-gate/.
  4. Gibson, Caitlin. "Video Games Are More Addictive than Ever. This Is What Happens When Kids Can't Turn Them Off." The Washington Post. Accessed July 15, 2020. https://www.washingtonpost.com/sf/style/2016/12/07/video-games-are-more-addictive-than-ever-this-is-what-happens-when-kids-cant-turn-them-off/.
  5. Koepp, M., Gunn, R., Lawrence, A. et al. “Evidence for striatal dopamine release during a video game.” Nature 393 (1998): 266-268. doi:10.1038/30498.
  6. Morikka. "NemoBotPollDataset." Kaggle. June 17, 2020. Accessed June 29, 2020. https://www.kaggle.com/dataset/2b67c3f1741051c8948dfd158514a93975cedb00e7e0a2b77ebc94f57bd2a0be.
  7. OECD. “Policy Brief: Ten Steps to Equity in Education.” OECD Publishing, Paris (2008): 1-8. Accessed July 30, 2020. https://www.oecd.org/education/school/39989494.pdf.
  8. Spencer, J. N. "New Directions in Teaching Chemistry: A Philosophical and Pedagogical Basis." Journal of Chemical Education 76, no. 4 (1999): 566-569. doi:10.1021/ed076p566.
  9. University of Reading. "Personal Response Systems (PRS)." Engage in Assessment. Accessed July 27, 2020. https://www.reading.ac.uk/engageinassessment/using-technology/delivering-assessment-using-technology/eia-tech-delivery-PRS.aspx.
  10. "COVID-19 National Survey Dashboard." CAMH. Accessed August 3, 2020. https://www.camh.ca/en/health-info/mental-health-and-covid-19/covid-19-national-survey.
  11. "Definition of Flipped Learning." Flipped Learning Network Hub. January 18, 2019. Accessed August 3, 2020. https://flippedlearning.org/definition-of-flipped-learning/.
  12. "GAMIFICATION: Meaning in the Cambridge English Dictionary." Cambridge Dictionary. Accessed July 13, 2020. https://dictionary.cambridge.org/dictionary/english/gamification.