Beyond the black box: how DeepSeek could transform higher education

DeepSeek, developed by a Chinese tech company, has recently taken the AI community by storm. This AI model showcases exceptional reasoning capabilities at a fraction of the usual training costs, potentially reshaping the higher education landscape. The implications of DeepSeek’s features for universities, educators and students are profound, offering both opportunities and new challenges. Here, I will explore how DeepSeek is impacting education through its transparent reasoning process and cost-effective development model. We will examine its potential benefits for educators and students alike, while discussing future directions for integrating AI tools such as these into educational workflows.

DeepSeek’s reasoning process

The transparency in DeepSeek’s reasoning process, as demonstrated in tasks such as computing 24 out of 5, 5, 5, and 1, offers a significant shift from the often criticised “black box” nature of AI. By detailing each arithmetic step, while describing the process of considering and testing various approaches to a problem, DeepSeek acts much like a human tutor. This facilitates a deeper learning experience, allowing users to see beyond mere calculations to understand the thought process behind problem-solving. This transparency strengthens trust in AI’s capabilities while enabling constructive feedback on its reasoning methods, fostering improvements and adaptations in AI behaviour.

For educators, this comprehensive insight into AI’s problem-solving approach goes beyond traditional teaching methods. It allows them to deepen their understanding of the nature of the tasks and the critical thinking involved. This is particularly beneficial in designing a curriculum because it tests knowledge and, at the same time, emphasises the development of problem-solving skills. By observing how AI navigates different solutions, educators can better guide students to explore various approaches to challenges, enriching their learning experience and promoting more profound engagement with the subject matter.

• DeepSeek and shallow moats: what does it mean for higher education?
• ChatGPT and generative AI: 25 applications in teaching and assessment
• How can we teach AI literacy skills?

From an assessment perspective, this level of transparency can significantly hamper students’ attempts to misuse AI in order to cheat. By emphasising the problem-solving process instead of simply delivering answers, DeepSeek encourages more authentic engagement with learning material. This focus on understanding and applying logical steps dissuades students from using AI merely as a shortcut to solutions. It promotes genuine learning and comprehension. Consequently, educators can leverage this feature to develop assessment methods that require students to demonstrate their thought processes, and foster a learning environment that prioritises analytical skills and educational integrity.

Reduce costs and expanding access

DeepSeek’s development at a lower cost than similar technology challenges the view that high-quality AI requires vast computational resources. Advanced tools such as DeepSeek have the potential to extend their reach beyond affluent educational sectors, becoming accessible to a broader audience. In contrast to expensive AI solutions such as OpenAI’s DeepResearch, which demands a subscription fee of USD $200 (£158) per month, DeepSeek is a financially viable alternative, crucial for students and teachers in the developing world, where even a monthly fee of USD $10 to $20 (£8 to £16) can be prohibitive.

DeepSeek’s global accessibility, powered by its open-source model, provides a competitive alternative to restricted AI tools such as OpenAI’s ChatGPT, which operates on a closed-source framework and has excluded countries such as China and Russia from accessing its services. While DeepSeek has been banned from offering AI services in countries or regions such as Italy, Taiwan and Australia, thanks to its open-source nature, third-party developers can still build on DeepSeek’s source codes to provide fine-tuned large language models or AI applications. In this way, open-source AI like DeepSeek acts as an antidote to geopolitical turmoil and protectionism, promoting inclusivity and technological equity by empowering global communities to overcome barriers imposed by closed systems and restrictive policies.

Improve functionality through APIs and overcoming barriers

The web browser and keyboard interface is the most common way to interact with generative AI chatbots but the linear, text-based interactions may not fully leverage the tool’s capabilities in complex tasks.

However, the use of Application Programming Interfaces (APIs) could be game-changing for the functionality and relevance of AI in education. Instead of starting from scratch, APIs allow developers to access DeepSeek’s functionalities and create tailored solutions that align with specific educational frameworks and cultural contexts. This customisation meets the diverse needs of global educational communities and supports various learning and teaching styles. Researchers and teachers may also leverage APIs to process large amounts of qualitative data and teaching materials more efficiently, without repetitive copy-pasting.

Despite the potential of APIs, developers often face financial barriers that can limit their effective use of AI technologies, such as licensing fees and the costs of accessing cloud services where AI models run. DeepSeek’s open-source model counters these challenges by allowing developers to access and adapt the model’s source code directly. This reduces costs associated with AI development and makes its educational applications more sophisticated. Developers can modify the AI’s algorithms and training data to better suit local needs, which can then lead to further training of the model on specific educational content or teaching methodologies. This adaptability is beneficial in regions where resources are limited and in educational settings where AI’s knowledge about local context is crucial.

By leveraging DeepSeek’s open-source model and strategic use of APIs, educational institutions and developers can overcome many of the barriers we currently face, fostering a more inclusive and innovative educational landscape. This approach can democratise access to advanced AI tools and encourage local adaptation and innovation. It’s a pivotal development in the field of educational technology.

Practical advice for academia

1. Customise for innovation: Universities should encourage and support their IT and educational technology departments to explore and experiment with DeepSeek’s open-source capabilities. Customising AI tools to fit specific academic programmes or research initiatives can lead to more effective and engaging educational experiences.
   
   
2. Capacity-building and API strategy development: Invest in workshops and training sessions for faculty and technical staff to learn about AI integration and API usage. This knowledge will empower them to design and implement AI-driven solutions independently. Universities should develop API strategies at institutional levels that involve API life cycle management, API security, educational alignment and measuring educational value, as well as AI-empowered developer enablement. 
   
   
3. Learning from deep reasoning: Encourage teachers and students to engage with DeepSeek’s transparent reasoning process as a dynamic learning tool. By revealing the inner workings of its problem-solving approach, DeepSeek offers educators a chance to adopt techniques inspired by the machine learning community, where unpacking complex algorithms fosters innovation. Teachers can use this transparency to model critical thinking and adaptability, sparking richer classroom discussions and encouraging students to value deep understanding over surface-level answers. This approach cultivates a culture of analytical rigour and intellectual curiosity, bridging insights from AI development and transformative pedagogical practices.

The advent of DeepSeek has the potential to transform the educational landscape by decentralising the concentration of advanced AI technologies. Its low cost and open-source attributes challenge the dominance of major AI suppliers, fostering a more inclusive and diversified technological ecosystem. This shift makes cutting-edge AI tools more accessible to a broader audience, and encourages innovation and customisation through local developer engagement. The strategic use of APIs further enhances this effect, allowing for tailored solutions that meet the unique demands of various educational settings around the world. As a result, DeepSeek democratises educational technology, while powering a global movement towards more equitable and effective learning environments.

Acknowledgement: This resource was written with the assistance of DeepSeek, Perplexity, GPT-4-turbo, Grok-3 and Kimi.

Simon Wang is lecturer and innovation officer at the Language Centre, Hong Kong Baptist University.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.