Adaptive learning systems  based on the learning that happens with online resources involving virtual social networks in the world where “, virtual social networks -widely democratized nowadays- are everyday learning spaces. Individual or collective “autodidaxis” (self-learning)  is widespread in these new universes suggesting   the scarcity or even the nonexistence of social learning relationships. ”(Cyrot & Jeunesse, 2012).

 However, The Belgian site reveals certain societal advantages. Indeed,It asserts that thanks to the ESCO system, “a European multilingual classification system of skills, competences, certifications and professions” used by adaptive learning platform, users “from different European countries can exchange resources related to competences / skills without any ambiguity. “The site also declares that the platform can be used as an informal learning network. (” We tested for you! MyLK, a tool to manage and share your digital learning ” , 2018). These two assertions suggest that digital adaptive learning  promotes “social learning” and integrates a large part of people. The criticism that is often made to digital learning technology to isolate learners in their learning would then be irrelevant in many cases even if the systems are based on the individualisation of learning/training paths  through the tracking of learning data and their processing by an AI.

 Actually on  digital learning  platforms the principle of social learning is highly present because users can “influence” each other on the content they consult and the paths they follow because “The outside world is… the factor that has the more influence on behavior. ”(Fenouillet, 2017, p. 30). This position makes echos  with Philippe Carré’s assertion: “we always learn alone… but never without others” 1. Thus, “If we already knew that self-study, whether yesterday or today, is not a form of soloformation, if we also knew that any act of learning, including when it is self-learning, is a social act, the exploration presented here also reveals the social dimension of learning mediated by the computer in the context of the use of the internet and contemporary virtual social networks. So, sociability (with others) naturally coexists with self-instruction (by oneself) even in face-to-face situations with the machine. “(Cyrot & Jeunesse, 2012)

The case of MYLK project: 

To illustrate the impact of the use of tracking, AI and adaptive learning in adult education and social learning, we can analyze the feedback from the test group on the MyLK platform. The tests took place in France, Poland and the Netherlands. Users were made up of 58% working people, 37% students and 5% business representatives. All participants were adults.

It is interesting to understand the feedbacks from the users because the tracking of learning data to offer personalized content (adaptive learning) through data analysis (AI) are central objectives of the platform.

User feedbacks have shown that 54% of users say that MyLK promotes learning through data tracking. 89% of those respondents think that MyLK actually allows them to manage the educational content they visited online. The ability of MyLK to suggest personal content was recognized by 61% of respondents. (Chrząszcz & Grodecka, 2018)

These generally satisfactory answers are surely to be put into perspective with other parameters, in particular the fact MyLK promotes exchanges between users which introduces a significant human and social factor which can have its importance in the overall experience of users. Indeed, several researchers (Amadieu & Tricot, 2015, 2015; Zourou & Lamy, 2008) showed that peer-to-peer exchanges are an important element in online educational systems.

The Role Human mediation in adaptive learning systems: 

Some adaptive learning system -managed entirely by the algorithm- design individualized paths without human mediation. This refers to what Amidieu calls the concept of “intelligent tutor or artificial intelligence”. According to him, the adaptive learning devices that work best are those in which there is human mediation. For him “it is not the machine that carries the adaptation but  humans”. (Amadieu & Tricot, 2015, p. 55).

To be effective, adaptive learning should therefore combine artificial intelligence and human mediation. 

The diagram below shows an adaptive learning system incorporating human mediation.

Indeed, human intelligence cannot be reduced to a network of logical action. On the contrary, the characteristic of human intelligence is to be intimately linked to emotions. Professor Antonio Damasio, expert in neuroscience, psychology and neurology explains that it is the emotions that make it possible to give value to an action: “good”, “bad” or “indifferent”. They are the ones that allow us to make sensible decisions by taking into account not only the results of past experiences and decisions but also what we have felt about these experiences and decisions; emotions therefore have an impact on memory too (Grandguillaume & Piroux, 2004). He adds that “emotional processes […] are action programs associated with affects, and feelings […] are the mental perception of the states of the organism (including the resulting states of emotions)” (Damasio in de Castex, 2017). Thus, without a human organism, it is impossible to produce human emotions and without emotions, artificial intelligence cannot give values ​​to the information it processes. In March 2016, Microsoft launched the chatbot Tay (conversational agent) on the social network Twitter. Using current knowledge of natural language processing, the company trained Tay to have the personality of a 19-year-old. After approximately 24 hours after its deployment, Tay was removed from the platform. Indeed, the robot learned from its conversations with users and ended up posting sexist, racist, anti-Semitic and aggressive tweets: Evolution of the Tay chatbot in ten hours, 2016 (Beres, 2016). In this case, the artificial intelligence system has failed to assess the relevance of the content it has learned, which sums up the impossibility of artificial intelligence to evade human intelligence. Even if artificial intelligence is endowed with human emotions, there is still no so-called non-intrusive technology that would allow it to capture cognitive and metacognitive data from users, such as emotions in response to a proposal, the affective states that motivate users’ choices, the logic of cognitive operations performed by them, etc. For example, a user answers “4” to the question “1 + 2 =? “. How to interpret this answer? is it an error from a lack attention, or a metaphysical approach to algebra? Thus, the machine does not have the capacity to capture the sinuous nature of the human spirit, it cannot respond to it in an adapted way. 

However, it should be noted that artificial intelligence has made it possible to develop new educational methods that has proven their efficiency over time and that are more easily implemented for a conventional training organization. Adaptive learning remains the most innovative among those new teaching approaches. 


Amadieu, F., & Tricot, A. (2015). Apprendre avec le numérique : Mythes et réalités. Consulté à l’adresse

Beres, D. (2016). Microsoft Chat Bot Goes On Racist, Genocidal Twitter Rampage. Consulté 27 Decembre 2019, à l’adresse HuffPost website:

Chrząszcz, A., & Grodecka, K. (2018). WP4—D4.2 Report of evaluation. Consulté à l’adresse

Cyrot, P., & Jeunesse, C. (2012). Autoformation et réseaux virtuels. Consulté à l’adresse

de Castex, E. (2017, décembre 7). Antonio Damasio : l’intelligence humaine ne peut venir aux machines – Anthropotechnie. Consulté 27 Decembre 2019, à l’adresse Anthropotechnie website: 

Fenouillet, F. (2017). La motivation (3e édition corrigée et actualisée). Consulté à l’adresse

Grandguillaume, A., & Piroux, C. (2004). A. Damasio. L’erreur de Descartes (1995) ; Le sentiment même de soi (1999) ; Spinoza avait raison (2003). L’orientation scolaire et professionnelle, 33(3). Consulté à l’adresse http://
Zourou, K., & Lamy, M.-N. (2008). Social networked game dynamics in web 2.0 language learning communities. Alsic, 16(Vol. 16).