Agent-based simulation of the learning dissemination on a Project-Based Learning context considering the human aspects

1  Abstract — This work p resents an agent-based simulation (AB S) of the active le arning process in an Electrical Engineering course . In order to generate input data to the si mulation, an acti ve learning methodology developed especially for part-ti me degree courses, called Project-Based Learning Agile (P BL A ), has been proposed and i mplemented at the Regional University of Blumenau (FURB), Brazil. Through t he analysis of survey responses o btained over five consecutive semesters, using partial least squares path modeling (PLS -PM), it w as possible to generate data parameters to use as an input in a hybrid kind of agent-based simulation known as PLS agent. The simulation of the scenario suggests that the learning occur faster w hen the student ha s higher levels of humanist’s aspects as self-esteem, self-realization and cooperation. Index Terms — Education, S tatistical analysis, Knowledge transfer, Analytical models, Electrica l engineering I. I NTRODUCTI ON n a r eflection on the contextualizatio n of engineers in the 21st cen tury, [1] argues that en gineers of tomorrow, a nd even toda y's engineers, will have to face deep and ne w challenges. For the author, the ne w engineers will have to dea l with the s tress of each day co mpeting in a world of accelerating changes where the y will have to solve problems without preced ence of scop e and scale. T he rec ent and fast technological changes h ave resulted in transfor mations in several fields, and a new par adigm has d riven the world to the information age [2 ], [3] . Concepts r elated to the humanization of e ngineering became i mportant in preparing e ngineers for the ne w challenges they will face in the f orthco ming decades. In a survey done with e xperts, the Great Engineering Challenges were d efined for the co ming years, presented without o rder of importance: stor ing solar en ergy, providing fusion e nergy, developing of carbo n sequestration methods, managing t he nitrogen cycle, providing access to clean w ater, developing better m edicines, advancing in health informatics, safe Laio O. Seman is with th e Department of Electrical Engineering, Federal University o f Santa Catarina, F lorianópolis, Braz il. Romeu Hausmann is with the Depa rtment of Electrical Eng ineering of Regional Unive rsity of Blume nau, Blumenau, Brazil . Eduardo A. Beze rra is with the Department of Ele ctrical Engineering, UFSC, Brazil , and with LI RMM, Université de Mon tpellier, F rance. cyberspace, preventing n uclear terro r, r estoring and i mproving urban infrastructure, reverse engineering the brain, improving virtual realit y, advanced personalized learnin g, scie ntific discovery [1]. It is perceived t hat the ch allenges involve energy and sustainability, health care, and advances i n the human capac ity for self-knowled ge, all of them in terdisciplinary aspects [1]. Interdisciplinary is an i mportant factor in b uildi ng t he necessary advance s for new engineers, and it is necessar y to build relationships with t he natural, social, b ehavioral, computational and mathematical sciences. It is necessary for engineering as a whole to co mmunicate with t he other areas of kn owledge to provide context for the students, providing the engineer with a view o f t heir role in history a nd in the future [4]. To this ability to move beyond basic kno wledge and achieve a level of understan ding that allo ws the engineer to deal with new p roble ms in an innovati ve and creative way, [5] calls "adap tive knowledge," and i mplies that such a facto r must be the new goal to b e achieved by educators in engineering, considering t he human aspects of t he education. According to [6] the co mmunity has made significan t advances in conducti ng studies related to engineeri ng education and p roposing goals that include this "adaptive knowledge", however, it has been less ef ficient at figuring out how to achieve those goal s. In the view of [5 ] , three educational co mponents must be not o nly developed , b ut aligned to complement each o ther to ac hieve these obj ectives: curriculum, instruction and as sessment. In this conte xt, this article proposes the focus on instruction, and how it is affected by individual (self -esteem and self- realization) and social ( coop eration) student’s aspects. For this p urpose, an active methodology called Project - Based Learning Agile was used as a case stud y o ver 5 semesters to model the stud ents' relation to humanization using par tial least squares path modeling (PLS -PM), later used as inpu t d ata in an agent -based si mulation, a social simulatio n that allows to evaluate how individ uals act a nd interact with each other [7] . The objective of the social simulation r aised in this work is to perceive the d ifferences in the speed of learning of a group of agents through different levels of humanization a nd instruction quality level. The r emaining of the p aper is organized as follo ws. Sectio n II pr esents the background and characterization of t he Agent - based simulation of the learning dissemination on a Project - Based Learning context considering the human aspects Laio O. Seman, Rom eu Hausm ann, Eduardo A. Bezerra I 2 methodology, Section III presents the results of the survey data analysis, Sect ion I V p resents the age nt -based simulation and Section V prese nts the conclusions. II. B ACKGROU ND AND C HARACTERIZATI ON The pilot p roject took place in the Depart ment of Electrical Engineering and Teleco mmunications of the Regional University o f Blumenau - Brazil (FURB) during the first half of 20 14 an d had the involve ment of the following courses : Power Electronics a nd Control and Servo mechanisms. Since the i nitial i mplementation, the proj ect was al so ap plied in 2015/1, 2015 /2, 2016/1 and 2016/2. During the applications of the project the students were challenged with aspec ts r elated, b ut not li mited, to designin g a CC -CC co nverter and it’ s closed loo p co ntrol and projecting a photovoltaic batter y charger. Two mai n aspects were considered i mportant in the implementation of the proj ects: tea m’s for mation (integrati on among st udents) and differentiation ( the same pro ject for all teams, but with different req uirements to allo w experience exchanging). Students who were not in the two courses intersectio n se t could be divided into tw o groups: one group o f t hose wh o already attended one course and are attending the other one; and the seco nd group would be those who are attending one course and would atte nd the other in t he future . O nly the second could crea te a problem in t he project’s process. S o there was a reco mmendation that the gro up should have members of both courses to share experie nces a nd kno wledge . The idea was to mitigated the issue. A. Pro ject-Based Lea rning Agile The part time p rofile of the Electrical Engineerin g major of FURB led to t he creation of a ne w applicatio n model of Project-Based Learning. T his, co uld be adapt ed to the student's p rofile, being dynamic a nd adaptive. T hus, its development was based on the principles of Agile p ractices. The Agile manifesto [8] was used as a basis, which was adapted to b etter meet the expected req uirements of such project, crea ting a methodology ca lled P roject -Based Learning Agile (PBL A ), according to the following p rinciples: Individuals and intera ctions o ver processes and too ls Working simulation over comprehe nsive documentation Student collaborat ion over deadlines negotiation Responding to cha nge o ver following a plan In the adap ted manifesto (as i n the original), eve n i f there is value in the ite ms o n the rig ht hand side (those not in b old), the highest value is gi ven to items o n the left hand side (bold ones). The A gile m anifesto principles are im portant to gi ve the project a flexible content. It is important that a project that can quickly adj ust alon g the way , to meet poten tial di fficulties t hat arise during the progress o f the courses that for m the proj ect. The difference of this prop osal is its goal. It d oes not aim to make t he students to follow t he Agile methodo logy within the PBL, but to use the principles o f the manifesto for the creation of P BL. As a result, only the instructors h ave contact with Agile, making the method ology transparent to stude nts. III. PLS- PM AS P ARA METER M ODELING T OOL In order to assess the impact of the methodology for posterior use with t he agent -based simulation, a sur vey fi rst proposed in [1 4] was applied in the po st-i mplementation pha se of the pro jects. T he idea is to measure the accepta nce of PBL methodology and to identify possible skills developed by t he students during the proj ect. As a way to oper ationalize the survey for the purpose of this work - to identify t he elements that construct active learning i n E lectrical Engineerin g based on humanistic concepts - five di mensions were taken from the literature: • PB L (composed by questions o f prefix P x ) • Learning (co mposed by questi ons of prefix C x ) • Cooper ation (composed b y questions of prefix G x ) • Self-estee m (composed b y questions of prefix E x ) • Self-realizatio n (co mposed by questions of prefix R x ) With the end of the p roject, the data anal ysis of the survey was made so that i t was possi ble to assess the p roject’s i mpact on the st udents. After five ap plications of P BL A , 162 student s’ responses were collected . Table 1 presents the used questionnaire and also the total percentage o f a greement a nd disagreement for each q uestion (including the five applications), ranging from 1 - "I completely disagree" to 5 - "I agree co mpletely" to the first group and 1 - " Not satisfactor y" to 5 - " Very satis factory" fo r the remaining groups. A. Mod el Hypotheses During the model co nception, the classical ed ucation literature ha ve been conside red, and hypotheses ha ve been raised to understand the importance of the h uman a spect in the educational process. In order to measure the impact of the humanization o n the proposed methodolog y and, consequentl y, o n learning, the relevant dimension s of hum an relations (se lf -esteem, self- realization and coo peration) w ere grouped into a new dimension called "Hu manization" . The new created dimension is based on the literature regarding the huma nization of engineering ed ucation and the 21st century required skills [9] – [12]. Its cr eation considers intrapersonal and interpersonal skills, a nd seeks to understand how both influence stude nt training. 3 TABLE I P ROCESS ASSESSMENT SURVEY Manifests Frequency (%) Neg. (%) Indif. (%) Pos. (%) 1 2 3 4 5 P1 I woul d like to repeat this exper ience in other co urses. 3.09 6.79 11.1 32.1 46.3 9.88 11.11 78.4 P2 I considere d the interdisciplinary relation posit ive to my learning. 0.62 7.41 5.56 35.1 51.2 8.03 5.56 86.4 P3 The deadline for the project com pletion w as satisfactory . 4.32 13.5 24.0 33.3 23.4 17.9 24.07 56.7 P4 At the end of t he project, I fulfilled the go als that I pursued. 1.24 6.17 9.88 58.0 24.0 7.41 9.88 82.1 P5 I did not feel overwhel med with the real ization of PBL. 12.3 29.0 34.5 16.0 7.41 41.3 34.57 23.4 Assess how the P BL impacted ... C1 ...in your abili ty to solve pow er electronics problems. 2.47 4.94 19.1 51.2 18.5 7.41 19.14 69.7 C2 ...in your abili ty to solve control problems. 3.70 3.70 14.8 43.8 33.9 7.41 14.82 77.7 C3 ...in your abili ty to make engine ering decisions. 1.85 2.47 16.0 50.6 29.0 4.32 16.05 79.6 C4 ...in your abili ty to seek informat ion for yourse lf. 0.62 0.62 14.2 48.1 36.4 1.24 14.20 84.5 C5 ...in your abili ty to solve probl ems presented in cl ass. 1.24 3.09 19.1 46.9 29.6 4.32 19.14 76.5 By participat ing in the group .. . G1 ...I fel t that cooperation helped to develop new ideas. 1.85 5.56 12.3 43.2 36.4 7.41 12.35 79.6 G2 ...usually I recognize the skills o f my colleagues. 0.62 0 4.32 46.9 46.9 0.62 4.32 93.8 G3 ...I fel t that everybody collaborated in the searc h for solutions. 3.09 9.88 15.4 34.5 36.4 12.9 15.43 70.9 G4 ...I appreciate the union created betw een people . 0 2.47 9.26 37.6 50.0 2.47 9.26 87.6 G5 ...I I ncrease the esteem fo r my colleagues. 0 2.47 12.3 41.3 43.2 2.47 12.35 84.5 In general… E1 ...I fel t comfortable whe n I had to face unfore seen situations. 1.85 15.4 26.5 42.5 13.5 17.2 26.54 56.1 E2 ...I managed to m inimize the neg ative effe cts of adversity. 1.85 6.79 20.3 54.9 16.0 8.64 20.37 70.9 E3 ...I kept me bal anced facing stressf ul situations 3.70 8.03 25.3 42.5 19.1 11.7 25.31 61.7 E4 ...I am aw are of my intelle ctual abilities. 0 1.85 11.7 55.5 29.6 1.85 11.73 85.1 E5 ...I bel ieve I have skills to be s uccessful. 0.62 3.09 6.79 46.3 43.2 3.70 6.79 89.5 Therefore... R1 ...I have w illpower to accomplish my goals. 0 0.62 5.56 46.3 46.9 0.62 5.56 93.2 R2 ...I invol ved all my skills in the w ork that w as done. 1.85 1.85 17.9 50.0 27.7 3.70 17.90 77.7 R3 ...I feel accomplis hed as a stude nt 1.24 6.79 16.6 43.8 30.8 8.03 16.67 74.9 R4 ...I feel that ins tructors contrib uted to my development. 1.24 4.32 9.88 38.8 45.0 5.56 9.88 83.9 R5 ...I feel that ev ery year I improve my skills. 0 1.24 6.79 38.8 52.4 1.24 6.79 91.3 It is consider ed that humanization exists at a level of abstraction be yond those that b uilds the individuality and t he cooperatio n and serves, for the p urposes of this work, as a way of grouping skills non -technical [9], [11] . After the assu mption o f s uch higher order construct, it is possible to infer the h ypotheses regar ding the constructs defined in the survey [13 ]. Accord ing to [1 2], the hypotheses of quantitati ve ori gin are predictions made b y the researcher regarding the expected relationships between variables, and their co nfirmation depends on the statistical procedure employed by the researchers o n the populatio n of a study. With the objective of understanding the formation o f knowledge with the ap plication of a PBL met hodology, taking into account the humanization o f the process, h ypotheses to investigate these relations hips are suggested. The hypotheses formulated and their rationale are presented b elow: Hypothesis A: Huma nization is a co mmon factor o f se lf- esteem; This presuppo sition seeks to understa nd the que stion of the student's personal satisfaction with himself in the h umanistic aspect of teaching [14] – [16]. It is ass umed that individuality is an i mportant aspec t in t he training of the engineer, responsible for helpin g or not his learning pr ocess, suppor ted by aspect s r elated to sel f -esteem, self-actualization and e motional background o f the student. Hypothesis B: Humanization is a common factor of self- realization; This assumption examines the student's relationship with his or her tendenc y to develop their growth capacitie s [14], [17], [18]. T his hypothesis has another aspect concernin g the assumption of the importa nce of individualit y in t he formation of the engineer. Hypothesis C: Humanizati on is a co mmon factor of cooperatio n; This assumption i s b ased on the transversal co mpetences o f the 21st century, which value coo peration as an integral part o f the modern world [11], [19] . In addition, it is also based on the ideas of [20] and [21] , authors that address the fragmentation o f the world's e xisting knowledge, and the importance of integration for the society progress. Hypothesis D: Humanizatio n has a p ositive influe nce on PBL; Humanization as t he foundation of the P BL is based on the concepts presented b y [22], who cites the i mportance o f an engineer involved with huma nitarian and social aspects, who is integrally involved in the community in a manner that the knowledge he acquires is useful. This hypothesis also seek s in [10] his confirmatio n, an author that addresses the importance understanding t he student’s own role on the word b efore the learning. This aspect was also s hown i n t he Maslow’s pyramid [23] w ith the assumption that knowledge will only be acquired if all human needs are satisfied. For [12] the complementation of the 4 technical and human aspects i s f undamental on the traini ng of a student for the society. Hypothesis E: P BL positively influe nces learning; On the assumption that dea ls with the positive influence of learning i n t he P BL, it is pos sible to reso rt to all the authors that have already ap plied the m ethodo logy i n the Electrical Engineering context [1 3], [24 ] – [32]. Relating to these hypothe ses, it is sugges ted that the proposed and applied PB L has its r oots in a humanist basis, formed b y i ndividualit y (self -esteem and self-realizatio n) a nd by cooperation among students, and thus sustained b y this humanization, PB L ser ves as the b asis for learning. As shown in Figure 1 1 . Self -e st e em Self -real i za t io n H um aniz at i o n PBL Co o p e rati on Lea rni ng Fig. 1 . Analyzed model based on the literat ure hypotheses. C. Structural Eq uation Modeling The structural equation modeling is a multivariate anal ysis of seco nd ge neration which see ks to understand the relationship between two or more variable s simultaneously in order to assess the structura l composition o f the anal yzed aspect [29 ], [30] . The PLS-PM algorit hm (P artial Least Squares Pat h Modeling) was u sed in the modeling of structural equations. According to [31] , this al gorithm has proved to be ad equate when the a nalysis has an explo ratory aspect . In additio n, current studies show that the technique has been conso lidating increasingly in explorator y studies of social scie nces [30]. The h ypotheses ha ve been anal yzed u sing the P LS-PM algorithm presented in Algorit hm 1 , and ar e sho wn in T able I, defining the p ath coe fficient (linear regressio n of the scores ) between each connected latent variable. 1 Arrows conn ecting huma nization to se lf-esteem, self-realization a nd cooperation are in the opposite direct because th ey are considered all common effects of humaniza tion. ALGO RITHM I PLS- PM PLS Path Modeling with path schem e, standardized latent variable scores and OLS regressions [32]. Input: 1 [ , ..., , ..., ] qQ X X X X  , i.e. Q blocks of cent ered manifest variables; Output: q w , q  , j  ; 1: for all q = 1, ... , Q do 2: initialize q w 3: 1 q P q pq pq q q p v w x X w       4: ' 1 ' ' ( , ) ( ' ) qq qq q q q q co r v v e v v v v        5: ' '' '1 Q q qq q q ev     6: update q w , cov ( , ) pq pq q wx   7: end for 8: Steps 1-7 are repeated until conv ergence on the outer w eights is achieved, i.e. unt il: __ m ax{ , , } pq cur rent iter ation pq prev ios iret arion ww    where  is a converge nce tolerance usua lly set at 0.0001 or l ess 9: Upon converg ence: (1) for each blo ck the standardize d latent variable scores are computed as weighted aggre gates of manifest var iables: ' ˆ q q q Xw   (2) for each endog enous latent v ariable ( 1 , ..., ) j jJ   the vector of pat h coefficients is est imated by me ans of OL S regression as: 1 ' ˆ ˆ ˆ ˆ ( ' ) ' jj       where ˆ  includes the scores of th e latent variables that explain the j -th endogenous latent variable j  , and ˆ j  is the latent variable score of the j - th endogenous la tent variable . TABLE I P ROCESS ASSESSMENT SURVEY Hyphothesis Interaction Influence 2 H A Humanization → Se lf - esteem 0.874 H B Humanization → Se lf - realiz ation 0.866 H C Humanization → Cooperation 0.753 H E Humanization – PBL 0.733 H D PBL → Le arning 0. 725 The results were validated using t he methodology proposed by [33 ], evaluating the indicato rs reliability (indicators with loading bigger than 0,7), the internal consistence reliability (indicators of a same dimension share a high correlation) and the discri minant validity (indicators are better represented by the dimensions they were allo cated). Also the res ult was tested with a nd B ias-Correct and Accelerated bo otstrapping pr ocess [34]. After assessing t he results ge nerated b y the P LS-PM it was possible to validate Hu manization as a co mmon factor of the individual (sel f-esteem r elationship o f 0.874 and self- realization relationship o f 0. 866 ) an d social (cooperation o f 0.753 ) aspects. Humanization also had a high influence on PBL (0.733 ). Finally, P BL also showed significance influe nce on learning (0.72 5 ). 2 All value s had a p-value < 0,001. 5 IV. PLS AG ENT (PBL C LASSROO M M ODE L ) The agent-based simulation (ABM) is part of a class of computational models to simulate the actions and interacti ons of autonomous agents a s a way of analyzing t heir ro le as a whole. In the p roposition made by [7], P LS-PM can be used as a way of quantif ying t he cause -effect relationships between the studied p henomena and later b e used as input para meters in ABM. The r esults of the actions to which the age nts ar e subjec t are given b y pr obabilities. T he pro bability of an event to occur with a n agent is as a sum of the PLS -PM path coe fficients ( , ij  ) of the latent variable connected to that event divided b y the maximum expected sco re of event, as sho wn in (1) , 1 , 1 max( ) J ij j i J ij j P        (1) Thus, the interpretation of the agent on the studied effect i s based o n the di mensions p reviously d efined b y the survey a nd quantized b y the PLS -PM algorithm resulting p ath coefficients. In the pro posed model, the main effect to be analyzed is the learning, considering di fferent design p oints for the humanizatio n and fo r the educatio nal proce ss (PB L), in a model called “PB L Classroom Mod el”. The path coefficients used in the pr ocess are the total effects, defined even if the dimensions are not directly connected, i.e.: even if humanizatio n is not directl y connected on lear ning, it has an indire ct influence effect calc ulated as (2), where the arro ws points the direction o f the influence, so the total effect of hu manization on learning is 0,53 2. () ( ).( L ea rn ing) Hu m aniz atio n Le ar ning Hu m aniz atio n P BL PB L   (2) The p robability of humanization considers the influences o f self-esteem, self-realization and co operation. For each of the causes of humanization is as signed a maximum scor e (in t his case 10 as suggested by [3 4]) wh ich is multiplied by the total effect of ea ch latent variable, the same is done for learning, which depends on humanization and P BL. This way t he , 1 m ax( ) J ij j LVs core   is shown on T able II. TABLE II A GENT M AXIMUM M ODEL Dimension Maximum Score Total Effects Result s Hum anization Max Value Cooperation 10 0.753 7550 24950 Self-esteem 10 0.874 8740 Self-re alization 10 0.866 8660 Learning Max Value Humanization 10 0.532 5320 12570 PBL 10 0.725 7250 Having the maximu m scores calculated, probab ility of humanization and lear ning ar e calculated as s hown in (3) and (4). Note that the pr obability o f learning depends o n the probability of humanization. 0.8 74. ( - ) 0.8 66. ( - ) 0.75 3.( ) 24950 hum aniza tion self este em self rea lizati on co ope rati on P    (3) 0. 53 2 .( ) 0.7 2 5. ( ) 12570 lea rn ing hu m a n iza ti on P BL P   (4 ) The simulation p arameters are presented in T able III, with the op erator having to pr eviously de fine the experi mental factors values a nd the control variables to obtain the desired response variables. In the objective of this work, it is desired to disco ver the d iffusion factor of lear ning for d ifferent values of humanization and qualit y of the applied methodology. Following [7], [3 5] principles, the simulation investigation should follo w the 3k -factorial d esign. T his design requires that each experi mental factor has one low, one medium, and one high value. The simulation exper iments perform eac h possi ble combination based on these parameters. Both h umanization and P BL attributes have factor values of 2, 5, and 8. Existing parameters, th eir scales and ex peri mental d esign values a re shown in Table I II . TABLE I II E XPERIMENTAL D ESIGN P ARAMETER S Parameters Scales Experimental Desi gn Experimental Fa ctors Cooperation Є [0,…,10] (2, 5, 8) Self-estee m Є [0,…,10] (2, 5, 8) Self-re alization Є [0,…,10] (2, 5, 8) PBL Є [0,…,10] (2, 5, 8) Control variabl es Agents numbe r N 961 Link-chance Є [0,…,1] random Є [0.3,…, 0.7] Response Vari able Diffusion rate Є [0,…,1 ] The simulation in based on a single grid net work model, where every grid position is occupied with and i mmobile agent. Every agent is allo wed to link with a maximum of 4 agents in it s adj acent ce lls ( Von Neumann topology) considering the li nk-chance p robab ility, a prob ability that two neighbors have of connecti ng. On ly one kind of agent is considered on th e model, the student agent. At every step of the simulation the student is able susceptible to t wo events: humanization and lear ning. During the simulation the following consider ation was made: age nts with high index of coo peration when connecting with a gents also high index of coop eration, can shar e knowledge, generating a possibility of acti vation of the learning event. The initial possibility of the agents to connect is given by the link-chance, considered rando m during the simulations ranging fro m 30% to 70%. The following rules gover n the simulation: 6 Agents: stude nts Attributes: lear ned; humanized On crea tion: random chance of creating bo nds to its classroom colleagues ( neighbors) Step: At eac h step of the simulation the s tudents decid e whether or not to change their attrib utes: • The student can learn i f he has e nough pos sibility and have not tried to lear n before on his own; • T he student can b e humanized if he has e nough possibility o f humanization and have not tried to be humanized before; • T he student ca n connect with its neighbors if both have succeeded i n the link-c hance possibility. Af ter the connectio n, if bo th st udents have e nough level of cooperation, they can transfer knowledge. The simulation sce nario is based on an active learnin g environment based on the previous d efined P roject Based Learning Agile. A. S imulation Resu lts The simulation was conducted using the Mesa P ython library [ 36] and following the steps described in [35] as validation of the model. The model i s illustrated at t he Figure 2 to a medium de sign point scenario (factors values of 5), where the squar es represents the st udents who have not suffered an event, small circles are students who only l earned ( blue) or o nly humanized (red) , and big circles are students who got both learned and humanized. Fig. 2 . Model si mulation for a medium design point scenar io. First an er ror variance analysis must be conducted in order to verify t he number of necessar y runs p er model setting. As proposed by [35 ] the variation overs over increased number of runs must be analyzed. For this analyses 3 settings were used : a lo w design point ( cooperation = 2; self-esteem = 2; self- realization = 2; P BL = 2. link-chance = 30 %), a middle desi gn point (cooperation = 5, self-e steem = 5; self -realization = 5; PBL = 5; lin k c hance = 5 0%) and a high d esign point (cooperation = 8, self-e steem = 8 , self -realizatio n = 8, PBL = 8, link-chance = 70 %). T he results are shown in T able IV. The low standard deviation (SD) and coefficient of variation (C V) change from 300 to 500 runs for all design points s hows t hat t he model tends to wards t he stabilizat ion and that it is adeq uate to p roceed with the simulatio ns using 300 runs. The defined number of runs is used on the posterio r simulations o f this paper. TABLE IV A GENT M AXIMUM M ODEL Design point Diffusion rate 1 run 50 runs 100 runs 300 runs 500 runs Low Mean 0.0 0.10 0.11 0.13 0.13 SD - 0.04 0.03 0.02 0.02 CV - 0.37 0.26 0.19 0.18 Medium Mean 0.28 0.44 0.47 0.51 0.51 SD - 0.11 0.13 0.10 0.10 CV - 0.26 0.27 0.21 0.21 High Mean 0.64 0.82 0.87 0.91 0.92 SD - 0.26 0.21 0.17 0.16 CV - 0.31 0.24 0.18 0.17 Having decid ed the num ber of necessar y runs, Table V contain eta squared e ffect sizes 2  of each experi mental factor on lear ning diffusion r ate. Results s hown that P BL has the higher impact on the d iffusion rate ( 2  =0.61), wh ile cooperatio n, self-esteem and self -realization, all bein g part of a high order construct (humanization) have lower i ndividual impacts. How ever, it is i mportant to notice that cooperation has an small ad vantage o n its effect size ( 2  =0.12) when co mpared to the individuals aspects ( 2  =0.08). TABLE V E TA S QUARED ( 2  ) E FFECT S IZE Dimension Cooperation Cooperation 0.12 Self-estee m 0.08 Self-re alization 0.08 PBL 0.61 A ccording to [35] a sensitivity anal ysis is an important step of the simulation to u nderstand the agents interactio ns. Figure 3 sho ws the variatio n of the diffusion rate of the knowledge for different le vels of li nk- ch ance, which is the sa me as raisi ng the p ossibility to student’s i nterchange kno wledge when t he humanization level is hig h enough. Figure 4 shows t he violin plot of learning diffusion rate for different link -chance levels ( 30% to 40%; 40% to 50%, 50% to 60) . From t he link-cha nce sensitive analys is it is po ssible to verify that t he maximum r ate of learning diffusion only occurs when students share acqu ired knowledge with each other. 7 Fig. 3 . Learning diffusion rate variation between 30 % and 70% of link-cha nce. Fi g. 4 . Learning diff usion rate for different link-chance le vels After, two more simulations were conducted . One regardin g the lear ning diffusion rate for different humanization le vels, allow the P BL levels to vary (2 , 5, 8). The results a re shown in Figure 5, and show that t he humanizatio n is directl y connected to the learning. Fig. 5 . Learning diffusion rat e for different humanizatio ns levels. The second one r egarding the learning diffusion rate f or different methodolog y qual ity levels ( PBL), allows the humanization to var y (2, 5 , 8) . The results are sho wn in Figure 6, it is p ossible to note that eve n with a high met hodology quality level, the diffusio n rate onl y reaches its maximum f or a small sample (when the humanization is at 8). Fig. 6 . Learning diffusion rate for different methodolo gy quality levels (PB L). IV. C ONCL USION This p aper proposed and evaluated an age nt-based simulation of the learni ng dissemi nation o n a Proj ect -Based Learning context considering the hu man aspects . To raise the input parameters of the si mulation, an acti ve learning called PBL A (Pr oject-Based Learning Agile) was applied on the Regional University of Blumenau (FU RB) during five consecutive se mesters and the students were invited to answer a survey on the end of the ap plication. The survey answer s were first submitted to a second generation multivariate analysis technique known of PLS -PM (partial least squares path modeling), which aims to identify relationships between non-observed variables. T he results in the form o f causal relatio nships were then used as inp ut parameters to the agent-ba sed si mulation, in a model cal led “PBL Classroo m Model”. In the simulation process, student s with a high index of cooperatio n were allowed to exchan ge knowled ge with ot hers students with also a high index of cooperation, increasing the learning diffusion rate. From the simulation results it was possible to verify that t he learning diffusion rates are directly related to the humanization (sel f-esteem, self -realization and cooperation) and to the methodology q uality. It was al so possible to verif y that the diffusio n rates achieve higher values when the cooperatio n b etween students is higher, sugg esti ng that even with a high quality teachi ng met hodology, it’s necessar y that the s tudents share knowledge between t hem to ach ieve the maximum learning. T he results pr esented in this study s uggest t hat humanization, and mainly coo peration, is an i mportant part in the educatio nal formation process of a ne w electrical engineer , and that to achieve a higher learning di ffusion rate , the teaching-learning relationship is better when also take s in to account the human nat ure. 8 R EFERENCES [1] C. M. Vest, “Context and Challenge for Twenty -First Century Engineering Education,” J. Eng. Educ. , vol. 97, no. 3, pp. 235 – 236, Jul. 2008. [2] J. Taylor, “Enginee ring the informatio n age,” IEE Rev. , vol. 44, no. 6 , pp. 250 – 252, Nov. 1998. [3] M. Castell s, “The Rise of the Netwo rk Society,” Massachusetts Blackwell Publ. , v ol. I, p. 594, 2010. [4] S . M. Malcom, “The human face of engineering,” J. Eng. Educ. , vol. 97, no. July , pp. 237 – 238, Jul. 2008. [5] J. W. Pellegrino, “Rethinking and Redesigning C urriculum, Instruction and Assessm ent: What Contemporary Research an d Theory Suggests,” A Pap. Comm. by Natl. Cent. Educ. Econ. New Comm. Ski. Am. Work. , no. Novembe r, pp. 1 – 15, 2006. [6] E. F. Redish and K. A. Smith, “ Looking Beyond Content: Skill Development for En gineer s,” J. Eng. Educ. , vol. 9 7, no. 3, pp. 295 – 307, Jul. 2008. [7] S . S chubring, I. L orscheid, M. Meyer, and C. M. R ingle, “The PLS agent: Predictive modeling with PLS-SEM and agent- based simulation,” J. Bus. Res. , vol. 69, no. 10, pp. 4604 – 4612, Oct. 2016. [8] M. F owler and J. Highsmith, “T he agile manifesto,” Softw. Dev. , 20 01. [9] J. J. D uderstadt, “Enginee ring for a Changing Worl d,” in Holistic Engineering Education , New York, NY: Springer New York, 2010, pp. 17 – 35. [10] E. Morin, “Seve n C omplex Lessons in Education for the Future: Education on t he Move,” p. 92, 2 001. [11] S. D. Sheppard, J. W. Pellegrino, and B. M. Olds, “On Becoming a 21 st Century Enginee r,” J. Eng. Educ. , vol . 97, no. 3, pp. 231 – 23 4, Jul. 2008. [12] J. B. Toro , Códigos da moder nidade: capacidades e competências mínimas p ara participação produtiva no século XXI . Por to A legre: Fundação Maurí cio Sirotsky So brinho, 1988. [13] L . O. Seman, G. Gomes, and R. Hausmann, “Statistical Analysis Using PLS of a Project- Based Learning Application in Electrical Engine ering,” IEEE Lat. Am. Tra ns. , vol. 14, no. 2, pp. 646 – 651, Fe b. 2016. [14] E. Morin, “O s setes saberes nece ssários à educação do futuro,” 2001. [15] D. Kahnema n, “Rápido e dev agar: duas formas de pensar,” 2012. [16] U. Neisser et al. , “I ntelligence: Knowns and unknow ns.,” Am. Psychol. , vol. 51, no. 2, pp. 77 – 10 1, 1996. [17] A. H. Maslow , “I ntrodução à psicologia do se r,” 1968. [18] R. Ry an and E. Deci, “I n trinsic and extrinsic mo tivations: Classic definitions and new directions,” Contemp. Educ. Psych ol. , 2000. [19] J. W . Pel legrino and M. L . Hilton, Educatio n for L ife a nd Wo rk: Developing Transfera ble Knowledge and Skills in the 21st Century . Washington, D .C.: National A cademies Press, 2012. [20] M. Ridley , “The rational optim ist.” 2010. [21] F. A. Hayek, “The use of know ledge in society .,” Am. Econ. Rev. , no. 35, 1945. [22] W. A. Bazzo, “Uma nova equação civilizatória x Problemas contemporâneo s da educação. Mensagem do coordenador,” 2016. [Online]. Av ailable: http://www .nepet.ufsc.br/. [Accessed: 13-Apr- 2016]. [23] A. Maslow, Toward a psychology o f being. Princeton: D Van Nostrand, 1962. [24] J. Kim, “An Ill -Structured PBL-Based Microprocessor Course Without Formal Laboratory ,” IEEE Trans. Educ. , vol . 55, no. 1 , pp. 145 – 153, Feb. 2012. [25] N. Hosseinzadeh and M. R. Hesamzadeh, “Applicatio n of Project -Base d Learning (PBL) to the Teaching of Electrical Power Systems Engineering,” IEEE Trans. Educ. , vol. 55, no. 4, pp. 4 95 – 501, N ov. 2012. [26] J. W. Thomas, “ A rev iew of research on projec t - based le arning,” 2000. [27] R. H. Chu, D. D.-C. L u, and S. Sathiakumar , “Project -Base d Lab Teaching for Pow er Electronics and Drives,” IEEE Trans. Educ. , vol. 51, no. 1, pp. 1 08 – 113, 2008. [28] W. Powel l, P. Powell , an d W. Weenk, “Project Led Engineering Education,” 20 03. [29] D. G. Lamar, P. F. Miaja, M. Arias, A. R odriguez, M. Rodriguez , and J. Sebastian, “A p roject -based learning approach t o teaching powe r electronics: Difficul ties in the application of Project-Based Learning in a subject of Switching- Mode Powe r S upplies,” in IEEE EDUCON 2 010 Conference , 2010, pp. 717 – 722. [30] L . E. M. Bracken bury, L. A. Plana, and J. Pepper, “Syste m - on -Chip Design and Implementation,” IEEE Trans. Educ. , vol. 53, no. 2, p p. 272 – 281, May 2 010. [31] E. Guzman Ramirez, I. Garcia, E. Guerrero , and C . Pacheco, “A tool for supporting the design of DC-DC converters through FPGA-based experime nts,” IEEE Lat. Am. Trans. , vol. 14, no. 1, pp. 289 – 296, Jan. 2016. [32] A. Kumar, S. Fernando, and R . C. Panicker, “Project -Based Learning in Embedded Systems Education Using an FPGA Platform,” IEEE Trans. Ed uc. , vol. 56, no. 4, pp. 407 – 415, No v. 2013. [33] J. F. Hair, M. Sarstedt, C. M. Ringle, and J. A. Mena, “An assessment of the use of partial least squares structur al equation modeling in marketing research,” J. Acad. Mark. Sci. , vol . 40, no. 3, pp. 414 – 433, May 2012. [34] B. Efron and R . J. Tibshirani, “A n Intro duction to the Bootstra p,” 1993. [35] I. Lorscheid, B. - O. Heine, and M. Meyer, “Opening the ‘black box’ of simulations: increased transparency and effective c ommunicat ion through the systemati c design of experiments,” Compu t. Math. Organ. Theory , vol. 18, no. 1, pp. 22 – 62, Mar. 20 12. [36] D. Masad and J. Kazil, “MESA: An Agent -Based Modeling Framewo rk,” Proc. 14th Python Sci. Conf. (SCIPY 2015) , no. Scipy, pp. 53 – 60, 2015. Laio Oriel Seman received the master degree in Electrical Engineering from the R egional Univ ersity of Blumenau in 2013. He is currently a PhD candidate at the Feder al University of Santa Catarina. His interests include multivariate da ta analysis and struc tural equation modeling. Romeu Hausm ann received the master and PhD degree i n Electrical Engineering from the Federal University of Santa Catarina in 2000 a nd 2011 respect ively. He is currently a professor at th e Regional University of Blumenau (FURB), focusing his rese arches in DC-DC and m ultilevel conv erters. Eduardo Augusto Bezerra PhD in Computer Engineering, University of Susse x, En gland, UK . He i s a lecturer at UFSC, Brazil, and visiting research er at LI RMM, France. His research i nterests include embedded systems for space applications, computer architect ure, and reconfig urable systems (F PGAs).