Patent Application
20220005367
The present invention relates generally to self-directed adaptive education.
Collegiate education is based on traditional lecturer-student interactions where the educator has a preset construct of how the course material should be conveyed to the students, usually in the form of a lecture, for a set amount of time, at frequent intervals, weekly or otherwise. Students are expected to learn the course material via the lectures as well as textbooks and other supplementary methods. Students can pose inquiries to the educator regarding the material to improve their understanding. As a basic educational structure this method works, but can inhibit both those who quickly grasp the subject matter and those who struggle, since they are all exposed to the same information at the same pace.
Much the same can be said regarding life-long learners, outside of a formal education program. Such students begin picking and choosing topics that interest them for professional, business, or personal interest.
Naturally, social connections form between students in similar curricula, professions, or businesses as they go through coursework together, as well as those that live together or participate in business or professional activities together. These connections grow into a social network of those participating in the common endeavor. From this social network, learning :styles can be observed and extended to groups of similar students to help them learn pricker and with greater impact. Adding this social network information to an adaptive learning system can allow better assignment and recommendation of content within the learning modules to each person, based on what the social network suggests about their interests, their preferred learning methods, and also the learning methods from their friends or colleagues. Using this educational aid, students can learn in a way that makes sense to them and lets them take more away from each education opportunity.
What is generally referred to as chunk learning entails recoding information into meaningful groups to be presented in a fashion to increase learning efficiency or capacity. The groups, called chunks, are formed based on meaningful or familiar relationships. Working memory capacity is increased by reducing the load presented to it. “In this way, the organism is able to decrease the amount of information that must be held in working memory by increasing the amount of information per chunk. Learning by chunking increases working memory capacity by reducing memory load and facilitates acquisition or recall by organizing long-term memory for information in perceived stimuli, motor sequences, or cognitive representations.” [Fountain S. B., Doyle K. E. (2012) Learning by Chunking. In: Seel N. M. reds) Encyclopedia of the Sciences of Learning. Springer, Boston, Mass.].
The subject technology is an improvement. The manner of forming and presenting CHUNKs can be improved to allow exploration, a networked rather than linear approach to learning, recommendation of learning pathways informed by social media information from other users, and dynamic adaptations based on the particular learner's progress.
What is needed is a system that allows each student to learn in ways that are effective for that particular student, whether it be watching videos, reading a book, reading through slides of the material, working example problems, running code, or a combination of those and other methods. Learning through these activities frees up the lecturing time, allowing the educator to teach at a higher level with deeper classroom discussion; whether that be critical thinking about the learned topics, teaching at an accelerated rate, focusing more on hands-on examples of the learned material, etc.
Embodiments described herein provide self-directed adaptive education. The adaptive education system includes a storage device to store aggregated learning content that includes learning units with learning material, the learning material having audio material, visual material, audiovisual material, or interactive material, CHUNKlets where each CHUNKlet is a CHUNKIet type and includes learning units, where the CHUNKlet type is an introductory type, an assessment type, an application type, or a methodology type, and CHUNKs where each CHUNK includes CHUNKlets. The adaptive education system also includes an aggregation engine to group learning units into CHUNKlets and CHUNKlets into CHUNKs based on inputs from course authors and define prerequisite relationships between CHUNKS based on inputs from course authors.
Embodiments in accordance with the invention are best understood by reference to the following detailed description when read in conjunction with the accompanying drawings.
Embodiments in accordance with the invention are further described herein with reference to the drawings.
The following description is provided to enable any person skilled in the art to use the invention and sets forth the best mode contemplated by the inventor for carrying out the invention. Various modifications, however, will remain readily apparent to those skilled in the art since the principles of the present invention are defined herein specifically to provide adaptive education.
Embodiments herein describe a real-time and adaptive teaching-learning method for enhanced and personalized education. They provide a curated way of moving through a Network of Knowledge composed of reusable learning objects joined together by common attributes (i.e., tagged with competency or skill levels), rather than following the standard linear or tree-like system of lectures or chapters. CHUNK Learning thus enables the learner to Heuristically discover or learn based on personal background and interests, which should not only enhance the learner's talents but make them a more valuable resource.
In CHUNK learning, a learner's interests determine his/her own learning path through the Network of Knowledge with individualized learning outcomes. Each student benefits differently from the learning experience, based on his/her skills and desires. Simultaneously, the Network of Knowledge builds on the experiences of the students covertly guiding learners through the educational materials, much like online retail stores provides recommendations for buyers. This can be achieved by moving away from interdisciplinary teaching that transfers methods from one discipline to another, opting instead for a trans-disciplinary teaching approach that crosses the boundaries of many disciplines using a diverse choice of teaching tools and software.
A “learner,” alternately called a “user” or “student” is an individual seeking knowledge through use of the system 100.
A “learner profile” comprises information from the learner regarding interests and preferred learning styles.
A “CHUNK” (Curated Heuristic Using a Network of Knowledge) comprises a topic to be learned, roughly equivalent to a section in a textbook. CHUNK content is broken down into smaller education materials, called “CHUNKlets.”
The “CHUNKlets” capture the breaking down of a topic into short and intense educational materials, allowing the learners to be engaged for a short period of time (and practice) before continuing to the next CHUNKlet. The CHUNKlets are categorized into four types: “Why”, “What”, “Methodology”, and “Assessment”. For each CHUNK, the CHUNKlets within the same category are interchangeable as they present the same topic from different points of view and different methods of delivery (e.g., video, audio, presentation slides, textual documents, etc.), allowing for personalized education when the most appropriate CHUNKS and/or CHUNKIets are suggested to the learner.
A many-to-many relationship exists between and among CHUNKs and CHUNKlets.
The subject technology uses a two-step process for recommending CHUNKs and CHUNKlets to each user to support personalized education, based on its current structure. First, relevant CHUNKs and CHUNKlets are determined based on the user's academic requirements and goals. Second, for each relevant CHUNK, the relevant CHUNKlets are ranked based on the user's learner profile and the social connections they share with other users in the system. The goal of this process is to maximize the chances that the learner will engage with CHUNKlets that are both useful and interesting to the learner. In one embodiment, the recommendation of a CHUNKlet is based on the learner profile's keywords. As such, personalizing the chosen CHUNKlet for each user is complemented by creating and utilizing a social network that ranks relevant CHUNKIets for each user. The newly proposed rating for each CHUNKlet is generated using the learner's profile and how that information links him/her to similar users, building on the CHUNKlet feedback provided from previous learners in the network. This maximizes the chances learners use methods that work for them. The built-in rating system is used to (1) collect data from users that have completed a CHUNK or CHUNKlet and (2) affect the ranking the CHUNK or CHUNKlet receives for other related users in the network, with the strength of that effect being determined by the strength of the individual's social connection with other users.
The system 100 requires identifying relevant connections between the users to accurately recommend appropriate new CHUNKlets to users; that is to say it relies on the overlaying social network that emerges between the users of the CHUNK Learning system. The subject technology discloses methods to generate a social network to inform system 100 by having the social network assign and modify a score of each CHUNKlet, to be used for recommendations to other users. The social network's nodes are the individual learner profiles and the edges (weighted and undirected) connect nodes with similar attributes. Attributes from each student profile are extracted and saved. Examples of such attributes are the current degree, branch of military service, previous degrees, and extracurricular interests.
Referring now to
Learner 102, upon seeking education through system 100, approaches a learner graphical user interface GUI 106 to make the request and to provide initial information for learner profile 120. The request is presented to selector 110, which in turn communicates with presentation engine 114. Presentation engine 114 communicates the request through learner preference feedback engine 112, and in turn to learner profile 120, seeking information associated with the particular learner 102 making the request. Cooperation and communication among learner profile 120 and a recommender engine 116 determine an initial set of CHUNKs and CHUNKlets to present to the learner 102. That determination is further communicated to a mapping engine 124 which makes the selected CHUNKs and CHUNKlets available to the presentation engine 114 for presentation. Actual presentation is accomplished to a desired presentation device 108, based on learner preference. The presentation device 108 may be that same device used with GUI 106, a separate similar electronic device, a video or audio device, or other device matching the methodology for the particular CHUNKlet.
From time to time during an education session, learner 102 may seek assessment, which is managed by performance feedback engine 118. If a learner 102 successfully completes an assessment, the successfully assessment is associated with the completed. CHUNK to verify that competency is achieved. Assessment information is retained and is available to the recommender engine 116, to be used in subsequent recommendations of CHUNKS and CHUNKlets.
The system 100 also addresses a cold start problem; establishing an initially useful state on first use by the learner 102. That is, how best to match a new user to material that fits his or her interests and learning style; particularly when little to no knowledge of the user's actual preferences is assumed. In system 100, and particularly managed via recommender engine 116, an assumption is that the learner profile information provided by the average user is incomplete, and it will be updated as the learner progresses through the CHUNKs, making it easier to b at that point. In particular, an assumption is that the directed learners will provide the least amount of information, combined with a further assumption that their motivation to provide information is the lowest.
Also consider a network cold-start problem; With little user data on-hand, how best to acquire useful information over time to identify emergent connections and apply collaborative filter methods? Putting in another way, how does the network improve its recommendations and internal connections through implicit or explicit feedback? System 100, in one embodiment, present a hybrid networked approach to overcome the cold-start problems.
Learners and content are treated as nodes on a network, and system 100 combines elements of content-mapping with syntactic sorting to determine a learner's initial location on this network. System 100 incorporates feedback and learning objective completion to update the user's location in the network of knowledge and then provide the user with recommendations to help guide the leaner through the network.
The social network can be created using learner profiles 120 as the nodes, and the attributes of those nodes as criteria to create edges. If two nodes have the same attribute, they can be connected by an edge in the social network. Attribute selection is limited to a predefined set of options to ensure uniform responses for a given category. This minimizes errors during data entry and ensures rank and designator selections correspond to the selected service. As the network grows, new categories and/or attributes may be added.
For this description, a set list of categories and attributes is created to generate a. usable network, as a subset of the CHUNK Learning system's 100 list of attributes. The selected categories have either a drop-down list of attributes for single selection or a multiple-choice list for attributes which may contain multiple items, such as extracurricular interests and classes. in one example, the categories can be the following:
1) Rank
2;) Service
3) Designator/MOS
4) Masters (Current Curriculum)
5) Major (Previous Degrees)
6) Extracurricular Interests
7) Classes.
The model focuses on the recommendation of CHUNKlets by the recommender engine 116 and assumes CHUNKs have already been selected for the learner 102 directed for the course he/she is enrolled in, since the CHUNKlets are interchangeable within their category. For this example, the model can be limited to 11 courses, with each course containing exactly one CHUNK and each CHUNK containing exactly three CHUNKlets.
In some cases, a social network can be generated, in part, based on users 102 and their initial ratings for each CHUNKlet. As new users 102 are introduced to the social network, they are connected to existing users 102 based on the attributes they select. Stronger or weaker connections between the users 102 can be determined based on similarity of their attributes.
The social network is created by determining how strongly each user 102 is connected to every other user 102. First, each category is weighed for importance to determine social connectivity. As an example, current degree may be given a weight of three and past degree may be given a weight of one, indicating connections made using a user's 102 current degree are three times more important than connections made using a user's 102 past degrees.
Next, for each category, the category's weight is used to form weighted edges between users if the users share an attribute in that category. Finally, these edges are added together to form the connections in the overall social network, where the weighted edge between each pair of users 102 determines how well-connected they are. Those skilled in the art will appreciate that the social network described is optional. The social network can be implemented to address the cold-start problem.
The effect of the social network on CHUNKlet recommendations is now explored. As new users 102 are introduced to the network and connected to existing users 102, the score of a CHUNKlet is updated for that user 102 and may result in different recommendations. These suggestions for CHUNKIets are based on the highest scored CHUNKlet in that category.
Though the method for constructing the edge weights in the social network remains the same, three methods for the edge weights can be used to determine CHUNKlet ratings. Let x be a new user 102, and y, z be existing users 102 in the network.
1) The linear method: the CHUNKlet's rating is proportional to the social edge weights. If the weight of the edge x, y is 5, and the weight of the edge is 10, then user z 102 have twice the impact that user y 102 has on the suggestions presented to x.
2) The exponential method: the impact a user 102 has on CHUNKlet ratings grow exponentially with their social weight.
3) The tier method: in this method, connections between users 102 are split into three tiers according to the social weight connecting them. Highly connected individuals fall into Tier 1, followed by Tier 2, and then Tier 3, as their social weight decreases. All individuals in the same tier have the same impact on CHUNKlet ratings—i.e. 6 for the top tier, 3 for the middle, and 1 for the bottom tier.
Each method has potential benefits and drawbacks. The tiered approach prevents highly connected users 102 from drowning out less connected users 102 but could also result in dissimilar users 102 having the same effect as those slightly similar, depending on the bounds of each tier. The exponential method does the opposite, it magnifies the effect highly similar users 102 have on each other. The linear method is the middle ground between tiered and exponential. As more users 102 interact with the CHUNKlets the recommendations of the recommender engine 116 become more robust.
A learner can use the network of knowledge 200 to quickly review their learning progress in CHUNKs 210 of various learning topics. In this example, the learner can identify potential CHUNKs 210 to complete next.
“Why”; Tantalizing the Learner 302: for example, learners open a “Why” CHUNKlet 302 to reveal an enticing one-of-its-kind educational trailer. The goal of CHUNK Learning is to make the student eager to learn, so CHL Ks can begin with a demonstration on why learning a particular topic is important. Much like a movie trailer attracts movie-goers to a movie, the “Why” CHUNKlet 302 attracts an exploratory learner to the CHUNK Learning module, answering the following questions:
Why is the topic relevant?
Why should students learn the topic?
“How”: Applications, Real and Relevant 304: learners dive into the “How” CHUNKIet 304 to uncover real and relevant applications. Here, learners discover the answer to the often-asked question, “When will I ever use this in real life?” Answers the following questions are also sought:
How is the topic applied in practice?
How does the learner validate what he/she already knows?
How are the learning outcomes tested?
How is new information, anchored to the learner's interests, incorporated into the module?
How can the learner apply the acquired skill/knowledge?
Methodology: A Variety of Delivery Methods 306: instructors carefully curate the Methodology CHUNKlets 306, guiding students though a variety of personalized course materials and delivery methods, including MOOCs and Creative Commons Licensed resources, as well as instructor-created content. For interactive modules, it is envisioned that instructors can follow the “I do it, We do it, You do it” model. The “Methodology” CHUNKlets' 306 main focus should be on answering the following questions:
What new information and skills will the module deliver?
What activities will the learner be required to perform?
What learning outcomes will the learner acquire?
What different methodologies could be used to engage with this new knowledge?
Assessment: Competency Based 308: learners can jump into the “Assessment” CHUNKlet 308 at any point to test their knowledge on any given topic. Assessments are available for every CHUNK. Opportunities for remedial learning may be present. Successful completion results in a CHUNK competency credit.
What is the competency-based framework, designed around learning objectives, needed for each CHUNK?
How should remediation be tested?
How should the post-test differ from the pre-test?
In block 404, each CHUNKlet can be tagged with learning outcomes (e.g., acquisition of target skills) and descriptive keywords. The tagging can be performed as described above with respect to block 402 t. e.g., course author, artificial intelligence, etc.).
In block 406, prerequisite relationships and disciplinary relationships are mapped between CHUNKs. In one example, the course author can specify the prerequisite and/or disciplinary relationships between CHUNKlets and/or CHUNKs. Disciplinary relationships may be intra-disciplinary or inter-disciplinary such that the same learning content can be used across multiple disciplines.
In block 4B, a network of knowledge can be generated based on the groupings, tags, and relationships. For example, the network of knowledge may be as described above with respect to
In block 424, a learner profile is created using the collected learner information. The learner profile can include learner preferences, learner attributes (e.g., degrees, certifications, skills, etc.), completed courses, etc. In block 426, relevant CHUNKs and/or CHUNKlets are selected based on the learner profile. In block 428, the relevant CHUNKs and/or CHUNKlets are used to build an exploratory path for the learner. In one example, the exploratory path may be similar to a portion of the network of knowledge described above with respect to
In block 446, feedback is obtained from the learner to update their interest level in topic and/or attributes of the learner. Similar as described above for
To make relevant recommendations in this example, the recommendation system relies on computing similarity values between pairwise CHUNKlets, the user and each CHUNK, and subsequently between the user and each CHUNKIet. To compute the similarity value, the cosine distance between two vectors in a 1×k-dimensional space or a 1×l-dimensional space, where k and l are the cardinalities of the network's 500 CHUNK or CHUNKlet keyword sets, respectively. The CHUNKs and/or CHUNKlets (across all CHUNKlet types) with the highest similarity value relative to the user are recommended first. Before providing a methodology for computing this similarity value, system information and structure requirements are outlined:
1) Initial System Inputs. The system resides in an information database, where each entity (CHUNK, CHUNKlet, and user) is identified with a profile(s). This profile has a unique identifier, a set of keywords, and, in the case of a CHUNK-CHUNKlet, a parent-child relationship. System administrators decide on CHUNK titles, and instructors upload CHUNKlets. When CHUNKlet upload occurs, the instructor must do four things: define the parent-child relationship between the CHUNKlet being uploaded and the CHUNK that it is assigned, categorize the CHUNKIet with one of the four categories “Why”, “What”, “Methodology”, or “Assessment”, assign to the CHUNKlet content keywords, and assign to the CHUNKlet learning method keywords (Video, PowerPoint, etc.).
2) User Profile Vectors. Two profile vectors will be built for each user: one based on content keywords that will be used for computing similarity values between the user and each CHUNK, and one based on learning method keywords that will be used for computing similarity values between the user and CHUNKlet. The first will be a 1×k-dimensional vector, where k is the cardinality of the network's content keyword set, and the second will be a 1×l-dimensional vector, l being the cardinality of the set comprising learning methods keywords. The system populates the user's vectors when the user initially creates his or her profile. It is a binary vector, where a one represents the user's interest in that keyword, and a zero represents no feedback or negative feedback in that keyword. The way the system obtains these keywords from the user during initial profile build is left to the current system administrators.
3) CHUNKlet Profile Vectors. CHUNKlets have two profile vectors: a 1×k-dimensional content keyword vector and a 1×k-dimensional learning method keyword vector. They are populated when the instructor uploads the CHUNKlet into the CHUNK Learning system based on that instructor's input.
4) CHUNK Profile Vector. Like the user's content keyword vector, the CHUNK's keyword vector is 1×k-dimensional, but it is not a binary vector, rather it is the sum of the vectors of its CHUNKlets. That is, the value associated with each keyword position in the vector will be based on the parent-child relationship between each CHUNK and CHUNKlet. The keywords associated with the CHUNKlet that the instructor tagged during upload will aggregate within the CHUNK, and this aggregated number will be the value for the keyword's position within the vector. Therefore, unlike the user's initial content keyword vector of ones or zeros, the CHUNK's keyword vector is not limited to a binary value.
Now that the system has its requisite information and appropriate vector lengths, the cosine distance between vectors can be computed so that the CHUNKlets with the highest cosine distance value can be provided as recommendations. This is performed in a two-round process.
Recommendation Round. Using the standard linear algebra cosine distance formula, the distance is computed between the user's keyword vector and all CHUNK keyword vectors. CHUNKs are then ranked from highest to lowest similarity value, and the first ranked CHUNK is recommended first. The user can accept or reject the CHUNK that is recommended, but this example focuses on users that accept the first recommendation. Once the user accesses the CHUNK, another cosine distance is calculated between the user's learning method vector and all CHUNKlets associated with the current CHUNK. The closest in CHUNKlets for each CHUNKlet type are recommended in decreasing order, where in represents the desired number of CHUNKlets shown based on system administrators' input.
User. Feedback Round. During this round, the user completes CHUNKlets within the current CHUNK. Implicit feedback, such as the length of videos watched, may be captured during this phase. Further, explicit feedback, which can be captured at the completion of each CHUNKlet and CHUNK, may also be captured.
In the CHUNKlet case, the user can be presented with a choice of rating the CHUNKlet as either a “like” or a “dislike”. The user's learning method profile vector will then be adjusted by multiplying a scalar value to the vector entry associated with the CHUNKlet type, expanded upon later below.
In the CHUNK case, the user will be presented with the same “dislike” or “like” question regarding the CHUNK as a whole, but if the user indicates positive feedback, a second feedback question will be asked. To support an adaptive CHUNK Learning system, this feedback round presents the user with the top three keywords (based on frequency) associated with the CHUNK and asks the user for either positive or negative feedback for each of the three keywords. The feedback collected will then impact the keywords attached to the CHUNK.
Lastly, to make the profiles adaptive, the user's profile vector will then be adjusted by multiplying a scalar value to the keyword(s) position in his or her content keyword vector. Additionally, if the user indicates positive feedback on any of the three keywords shown at the end of the CHUNK, and that keyword is not already represented in the user's keyword vector, a “1” value will be added to the user's keyword vector before the scalar is applied. This enables the user to prolong his or her exploration in the CHUNK Learning network by making it possible for related CHUNKs to be suggested to the user.
In one example, the “like” scalar value can be set to 1.05 and the “dislike” scalar value can be set to 0.01. These values can be adjusted depending on system administrator preference. Because of these updates, the CHUNK Learning system can be considered to have “dynamic profiles”, since each user's profile adjusts according to explicit feedback.
Upon completion of a CHUNK, that CHUNK's similarity value to the user profile will be assigned the value zero. This is to prevent the user from being recommended a CHUNK that has already been completed.
The process then repeats. It should be noted that this methodology is applicable to both directed and exploratory learners. For the directed learner case, users may take a different path through the network than a purely exploratory learner might, but they can still use and benefit from the feedback mechanisms built into the system particularly in respect to the learning methods presented over time.
The invention may be implemented on virtually any type of computer regardless of the platform being used. For example, a computer system can include a processor, associated memory, a storage device, and numerous other elements and functionalities typical of today's computers. The computer may also include input means, such as a keyboard and a mouse, and output means, such as a display or monitor. The computer system may be connected to a local area network (LAN) or a wide area network (e.g., the Internet) via a network interface connection. Those skilled in the art will appreciate that these input and output means may take other forms.
Further, those skilled in the art will appreciate that one or more elements of the computer system may be located at a remote location and connected to the other elements over a network. Further, the invention may be implemented on a distributed system having several nodes, where each portion of the invention may be located on a different node within the distributed system. In one embodiment of the invention, the node corresponds to a computer system. Alternatively, the node may correspond to a processor with associated physical memory. The node may alternatively correspond to a processor with shared memory and/or resources. Further, software instructions to perform embodiments of the invention may be stored on a computer readable medium such as a compact disc (CD), a diskette, a tape, a file, or any other computer readable storage device.
This description provides exemplary embodiments of the present invention. The scope of the present invention is not limited by these exemplary embodiments. Numerous variations, whether explicitly provided for by the specification or implied by the specification or not, may be implemented by one of skill in the art in view of this disclosure.
It is to be understood that the above-described arrangements are only illustrative of the application of the principles of the present invention, and it is not intended to be exhaustive or limit the invention to the precise form disclosed. Numerous modifications and alternative arrangements may be devised by those skilled in the art in light of the above teachings without departing from the spirit and scope of the present invention.
This patent application is a non-provisional of and claims the benefit of U.S. Provisional application No. 63/047,987, filed Jul. 3, 2020 which is hereby incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 63047987 | Jul 2020 | US |
