161
C i ical hinking and a i icial
in elligence in academia: A
quali a i e ma ix analysis p ocedu e
o e alua ing AI sys ems
Lluís Codina
Uni e si a Pompeu Fab a, Spain
h ps://o cid.o g/0000-0001-7020-1631
Elisenda Aguile a-Co a
Uni e si a Pompeu Fab a, Spain
h ps://o cid.o g/0000-0003-0923-9192
Ca los Lopezosa
Uni e si a de Ba celona, Spain
h ps://o cid.o g/0000-0001-8619-2194
Pe e F eixa
Uni e si a Pompeu Fab a, Spain
h ps://o cid.o g/0000-0002-9199-1270
Codina, L., Aguile a-Co a, E., Lopezosa, C., & F eixa, P. (2025). C i ical hinking and
a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI
sys ems. In J. Gualla , M. Vállez, & A. Ven u a-Cisquella (Coo ds). Digi al communica ion.
T ends and good p ac ices (pp. 161-173). Ediciones P o esionales de la In o mación.
h ps://doi.o g/10.3145/cu icom.12.eng
162
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Abs ac
This wo k in oduces he Ma ix AI Sys ems Analysis P ocedu e (MASIA), a quali a i e, ma-
ix-based me hod designed o e alua e he pe o mance and quali y o gene a i e a i icial
in elligence sys ems wi hin academic se ings. MASIA cen e s on he analysis o h ee key
componen s in AI-gene a ed esponses: na a i e syn hesis, sou ce usage, and he o mula-
ion o new p omp s. By doing so, i os e s c i ical hinking among use s and o e s aluable
ools o bo h eaching and esea ch.The p ocedu e de ines a iables and analy ical pa am-
e e s ha enable he compa ison o di e en AI sys ems, he eby suppo ing in o med deci-
sion-making in schola ly and esea ch en i onmen s. Fu he mo e, MASIA in eg a es e hical
conside a ions, including aceabili y, p ope a ibu ion, and plagia ism p e en ion, making
i a lexible ins umen adap able o a ious academic needs and p ojec s. The chap e con-
cludes ha MASIA is a s aigh o wa d ye powe ul ool o enhancing c i ical hinking, im-
p o ing eaching and lea ning p ocesses, and p o iding a ounda ion o compa a i e e-
sea ch on a i icial in elligence in academia.
Keywo ds
Gene a i e a i icial in elligence; Quali a i e e alua ion; C i ical hinking; Analysis ma ices;
Academic e hics; AI sys ems in academy; E alua i e me hods.
1. In oduc ion
This pape p esen s an analy ical p ocedu e o e alua e he pe o mance and quali y o
gene a i e a i icial in elligence sys ems in academic en i onmen s.
The p ocedu e, which we call he Ma ix AI Sys ems Analysis P ocedu e o MASIA, is designed
o e alua e AI sys ems ha , as pa o hei esponse, no only p o ide a na a i e summa y
bu also include ci a ions and he bibliog aphic sou ces hey used o gene a e he con en .
This me hod o analysis p omo es c i ical hinking among AI sys em use s, p o ides elemen s
o eaching-lea ning p ocesses, and can be he basis o de eloping da a collec ion in e-
sea ch p ocesses.
The use o sou ces as pa o he esponse is a necessi y in he academic con ex because one
can e i y and expand he in o ma ion p o ided by he AI, as well as main ain he chain o
a ibu ions (High- Le el Expe G oup on A i icial In elligence . 2019; C omp on and Bu ke,
2023¸ Kaebnick e al. 2023; Lund e al, 2023; Tilie e al., 2023; Gunde sen e al., 2018; Wo ld
Commission on he E hics o Scien i ic Knowledge and Technology, 2019; Dwi edi e al.,
2021; Bianchini e al. 2022). The la e is doubly con enien , because in addi ion o inc easing
he quali y o he AI esponse, i p e en s plagia ism o inadequa e a ibu ion, bo h o which
a e essen ial in academic wo k.
The p ocedu e p esen ed he e consis s o analysis ma ices based on a se ies o a iables (Co-
dina & Ped aza, 2016), which a e de e mined acco ding o he au ho s’ eaching expe ience
and p io expe ience in analyzing and using AI sys ems o academic wo k, pa icula ly due
o he need o p o ide p o ocols o he c i ical use o AI sys ems o uni e si y s uden s and
p edoc o al esea che s.
163
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Since he launch o Cha GPT in la e 2022, he au ho s ha e inco po a ed a i icial in elligence
in o hei eaching and esea ch ac i i ies (Lopezosa & Codina, 2023; Lopezosa e al., 2023a;
2023b; Aguile a-Co a e al., 2024a; 2024b; Codina, 2025). This in eg a ion highligh ed he
need o an in ellec ual ool o ain s uden s and p edoc o al esea che s in he p ope use o
AI (Codina and Ga de, 2023). The e was also a need o an ins umen ha would allow com-
pa a i e s udies o he e iciency o a i icial in elligence sys ems sui able o use in academia.
The la e could be use ul o esea ch pu poses, o o p o iding economic decision-make s
a uni e si ies, o example, wi h in o ma ion on which o e alua e he acquisi ion o AI sys-
ems (Bha ia, 2023; Whi ield & Ho mann, 2024; Else ie 2024).
Be o e p esen ing he componen s o he e alua ion me hod, we mus p esen some e mino-
logical cla i ica ions, he conside a ion o which is pa o he p ocedu e i sel , jus as we mus
conside he composi ion o he esul s o AI when i esponds o a use ins uc ion.
2. Te minology
The e minology p esen ed in he able below is conside ed pa o he p ocedu e, so i is
necessa y o p ecisely es ablish he use o a se o e ms o i s p ope applica ion.
Table 1
Te minology o he AI sys ems e alua ion p ocedu e
Te m Explana ion
Bibliog aphic sou ces
In a RAG- ype AI sys em (see de ini ion below), his is he lis o documen s (jou nal a icles,
epo s, web pages, e c.) ha jus i y he answe . In he con ex o he e alua ion p ocedu e,
his concep o sou ce is he one aken in o accoun unless o he wise indica ed.
Sou ces o in o ma ion
In a RAG- ype AI sys em (see de ini ion below), in o ma ion sou ces a e he esou ces
used o loca e he sou ces on which i s esponse is based. Typical in o ma ion sou ces
o RAG- ype AI sys ems can be academic da abases o sea ch engines like Google.
Indica o s
In an e alua ion p ocedu e, indica o s a e cha ac e is ics ha p o ide in o ma ion abou
wha is o be e alua ed o compa ed. Fo example, in a compa a i e analysis o na ional
economies, he unemploymen a e, in la ion, o GDP a e indica o s. By hei e y na-
u e, hey a e also a iables.
Ma ices
Table-based in o ma ion s uc u es ha allow da a o be ex ac ed, p esen ed isually,
and compa a i ely analysed. In he p ocedu e p esen ed he e, he use o ma ices is
conside ed no ma i e. The wo d “ able” is equi alen . The use o he e m “ma ix”
in oduces he idea ha a able mee s ce ain condi ions, he mos impo an o which
a e ha hey a e homogeneous ables and a e o ganized so ha ows a e en i ies and
columns a e he p ope ies o he en i ies.
AI model
Also called La ge Language Model (LLM). This is he echnical name o gene a i e
a i icial in elligence, gi en i s echnological ounda ion. An AI model o LLM canno be
used by an end use , as i s use equi es p og amming and APIs. The e o e, end use s
gene ally wo k wi h AI h ough AI sys ems.
Resul s page
An AI sys em’s esponse is p esen ed o he use on a esul s page ha ypically con-
sis s o a leas h ee componen s: he na a i e summa y, a lis o sou ces, and a se o
sugges ed new p omp s.
Pa ame e s
In an e alua ion p ocedu e, pa ame e s g oup indica o s o a iables. The idea is ha a
g oup o a iables se es o cha ac e ize a signi ican aspec o a ce ain complexi y o he
subjec being e alua ed. Fo example, one pa ame e o na ions in compa a i e analyses
is hei economy, ano he hei demog aphics o poli ical sys em, e c. Bu o cha ac e ize
each pa ame e , i is necessa y o use disagg ega ed indica o s, such as he unemploy-
men a e in he case o he economy, along wi h o he s such as GDP, e c.
P omp Na u al language ins uc ion used o ob ain he esponse om an AI sys em.
164
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Te m Explana ion
Sugges ed P omp s Lis o new p omp s ha some AI sys ems p o ide as pa o hei esponse o a p omp .
Rapid e iew
A ype o e iew ha omi s some o he usual con ols o sys ema ic e iews o ob ain
immedia e esul s wi h p elimina y alue. Na a i e syn heses om AI can be compa ed
o a o m o apid e iew.
Re ie al Augmen ed
Gene a ion (RAG)
Augmen ed Re ie al (ARG) in ol es imp o ing he answe s gene a ed by AI sys ems
by combining hei aining base knowledge wi h in o ma ion e ie ed in eal ime om
ex e nal sou ces such as academic o specialized da abases o gene al-pu pose sea ch
engines like Google o Bing. Mos academic AI sys ems a e ARG. Some gene al-pu pose
AI sys ems, like Pe plexi y, a e also ARG- ype sys ems. Google, p esumably, should be-
come ARG- ype once i e ec i ely in eg a es i s sea ch engine wi h i s AI.
Li e a u e e iew
A li e a u e e iew is a sys ema ic p ocess o sea ching, selec ing, analyzing, and syn he-
sizing exis ing in o ma ion on a speci ic opic. I in ol es c i ically e alua ing p e ious
s udies, iden i ying pa e ns, deba es, and gaps in cu en knowledge, and p esen ing
a comp ehensi e and o ganized o e iew o he s a e o he a in he ield. , P oducing
li e a u e e iews applied o academia is one o AI’s main unc ions.
Sc a chpad
The sc a chpad is he sec ion p eceding he na a i e syn hesis whe e he AI p esen s
he chain o easoning i ollowed o accomplish i s asks. In ou case, he ask consis s
o sol ing he ou phases leading o a na a i e syn hesis. The abili y o examine he
sc a chpad means, among o he hings, checking whe he he AI unde s ood he objec-
i es o he ask and co ec ly app oached i .
Na a i e syn hesis
A na a i e syn hesis is he esul o analyzing a se o sou ces using a well-de ined
amewo k and compiling he key insigh s de i ed om he analysis o hese documen s
in o a cohe en ex ual summa y. A na a i e syn hesis is bo h a byp oduc o a li e a u e
e iew and pa o he esponse o a gene a i e AI sys em.
AI sys em
I is composed o one o mo e AI models (also called LLMs o La ge Language Model),
o one o mo e so wa e laye s, e.g. , o que ying da abases (e.g. , academic da abas-
es), o managing e e ences, e c., as well as a use in e ace.
U ili ies
These consis o complemen a y unc ions ha so wa e p og ams ypically o e in ad-
di ion o hei co e unc ions. Fo example, in an AI sys em ocused on academia, a
u ili y migh consis o esou ces o managing e e ences.
Va iables
A a iable is a p ope y o an en i y ha can ake on di e en alues o each en i y, o
o e ime wi hin he same en i y. Reco ding hese alues allows en i ies o be analyzed,
cha ac e ized, and compa ed. They can also be called indica o s i hei heu is ic na u e
is desi ed.
Sou ce: own elabo a ion
3. Composi ion o an AI sys em’s esul s page
In a esul s page o he ype o AI sys em we a e analyzing, we can de e mine he exis ence
o h ee main componen s:
1. Na a i e syn hesis.
2. Sou ces consul ed.
3. New p omp s.
We will p esen each o hese componen s o e alua ion pu poses. Bu i s , we mus poin
ou a ou h elemen ha , al hough no pa o he esul s page, can be ound wi hin he in-
e ace i sel :
4. Addi ional u ili ies and unc ionali ies speci ic o each sys em.
165
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
3.1. Na a i e syn hesis
Na a i e syn hesis is he ex gene a ed by gene a i e a i icial in elligence I ’s a syn hesis
because i ’s he esul o analyzing and syn hesizing a se ies o p e ious pieces o in o ma ion.
I ’s na a i e because i ’s p esen ed as a mo e o less a icula ed na a i e o discou se. O he
ypes o syn heses a e possible, such as hose p esen ed in he o m o ables o g aphs.
Howe e , when we alk abou na a i e syn heses, we conside any o ma o be included, by
ex ension, unless o he wise s a ed.
F om he e, we can es ablish a i s block o c i e ia acco ding o which AIs ha p esen syn-
hesis a e p e e able, which a e:
– T aceable h ough a isible sc a chpad. The sc a chpad is a sec ion p eceding he na a-
i e summa y in which he AI anspa en ly and aceably p esen s he chain o easoning
i ollowed o sol e he ask. The abili y o examine his chain o easoning allows o he
de ec ion o po en ial bias o o he e o s, as well as e i ying whe he he AI co ec ly
unde s ood he ask. In any case, i acili a es he aceabili y o he p ocess ollowed.
– A icula ed. This means ha he summa y is p esen ed in some so o s uc u e. Fo
example, in sepa a e sec ions, possibly o ganized by headings and ollowing some so
o logical a angemen o sec ions.
– They exhibi cohe ence and cohesion. Cohe ence consis s o he in e ela ionship o
he sen ences ha make up he ex h ough hei app op ia e connec ion o he main
heme. I is mani es ed by hema ic uni y, he ( ela i e) absence o edundancy, and
he logical p og ession o ideas. Cohesion, on he o he hand, is mani es ed in he
g amma ical in e ela ionship o sen ences. I is p ima ily de e mined by he connec i es.
– They exhibi connec i i y. The end o each pa ag aph an icipa es he nex , and he
beginnings o subsequen pa ag aphs connec o he p e ious ones. This p ope y is
made e iden h ough he use o connec i es. Connec i i y is inc eased i he e is a
sec ion ha euni es he main ideas, o a sec ion wi h an equi alen unc ion.
– They a e ( ela i ely) long. All o he c i e ia being equal, long summa ies a e p e e able.
Since we’ e alking abou a ange ha can ex end om 200 o 3,000 wo ds, hose ha ,
i necessa y, can be close o his highe limi a e p e e able.
– They a e mul imodal. In addi ion o ex , hey include some addi ional o ma ing, e.g.,
ables, ca ds, concep maps, mind maps, o diag ams.
3.2. Sou ces
In academic AI, sou ces a e ypically documen s, epo s, and scien i ic jou nal a icles. Ideally
hey allow he ideas and con en comp ising he syn hesis o be a ibu ed o hei o iginal
c ea o s. In he academic con ex , we ha e a ca ego ical impe a i e o use AI sys ems ha ,
along wi h he gene a ed na a i e syn hesis, p o ide he sou ces on which hey a e based.
This is he eason o p e e ing RAG- ype AI sys ems.
An addi ional eason o his p e e ence is ha an absence o sou ces in he answe would lead
o a b eak in he a ibu ion chain and, consequen ly, p edispose o plagia ism. No e ha we
a e sepa a ing he ac ual ac (o lack he eo ) o plagia ism om he ac ha some AI sys ems
acili a e o p omo e, de ac o, plagia ism by o e ing unsou ced answe s. Logically, we should
p e e AI sys ems ha , a he e y leas , do no p omo e plagia ism.
166
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Fo he pu poses o ou p oposed analysis, AIs a e p e e able [ o wha ?] in ela ion o sou ces:
– They exhibi capilla i y. AIs ha assign sou ces a he sen ence le el, o ailing ha , a
he pa ag aph le el, a e p e e able o a inal lis ha a ec s he en i e undi e en ia ed
na a i e syn hesis. Capilla i y also implies connec i i y, since when a pa ag aph has
( o example) h ee ela ed ideas, each sou ce is linked o each o he ideas, ins ead
o placing he h ee sou ces a he end o he pa ag aph o a he end o he en i e
syn hesis.
– They p o ide well- o med ci a ion o ma s, ha is, wi h comple e e e ence in o ma ion
and, whe e app op ia e, iable links.
Fu he mo e, he use o AI is obliged o e i y and e iew sou ces, no only o e alua e a -
gumen s, bu also o a ibu e hi d-pa y ideas and con en o hei ue au ho s h ough he
con en ional ci a ion sys em.
3.3. New p omp s
Some AIs o e , as a hi d p ominen componen o hei esponses, a lis o new p omp s o
new ques ions. This app oach may be o li le in e es o may be e y incisi e. In he la e
case, hey ha e ob ious heu is ic alue. P e e able a e hose ha gene a e addi ional, ela ed
p omp s o ques ions as pa o he esul s page, which we will e alua e:
– Oppo uni y. Tha is, a e he new sugges ed p omp s app op ia e o he sea ch objec-
i es?
– Va ie y o app oaches. Do hey o e new ace s o app oaches no conside ed in he
o iginal p omp ?
3.4. U ili ies and Idio unc ions
Some AI sys ems p esen one o mo e cha ac e is ics ha a e speci ic and unique o he sys-
em unde conside a ion. In con as o common unc ions, we can speak o idio unc ions; ha
is, unc ions unique o each pa icula sys em and he e o e p esen only in he sys em unde
conside a ion. Fo example, an AI sys em may p esen a unc ion ha consis s o ex ac ing
concep s, o ano he ha consis s o being able o design analysis ma ices om e e ences.
These a e called idio unc ions because hey a e unique o each AI. Al hough hese unc ions
may become s anda dized o e ime (and he concep may lose i s meaning), hese di e enc-
es a e signi ican a p esen and use ul o conside .
3.5. Analysis ma ices
Wi h he help o he p e ious concep s, we can now p esen he elemen s o analysis, which
we a icula e in pa ame e s and a iables as shown in Table 2.
167
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Table 2
Analysis a iables.
Pa ame e Code Va iables / Check Ques ion
1. Na a i e
syn hesis
1.1
Sc a chpad
Does i p esen a sc a chpad wi h he chain o easoning ollowed by he AI sys-
em? Can his documen be consul ed a e he ask is comple ed?
1.2
A icula ion
Is he na a i e syn hesis p esen ed o ganized o a icula ed in a ious sec ions o
is i p esen ed as a con inuum wi hou a de ined s uc u e?
1.3
Cohe ence and connec ion
Is he e a consis en hema ic uni y h oughou he na a i e summa y and wi hin
each pa ag aph? Is he e a connec ion be ween he sec ions, pa ag aphs, o sec-
ions o he na a i e summa y?
1.4
Ex ension
Is he summa y o he na a i e adequa e o i s objec i es? How many wo ds does
he na a i e summa y con ain?
A e he e al e na i e e sions o he ex ension?
1.5
Mul imodali y
Does he esul s page include only ex , o does i include o he o ms o in o ma-
ion, such as diag ams? I no ini ially p esen ed, a e hey o e ed as al e na i es?
2. Sou ces
2.1 Numbe
How many sou ces a e ci ed?
2.2
Di e si y
Do he sou ces exhibi adequa e di e si y o he pu poses o he p omp ? No e:
A single da abase is no an a p io i limi a ion on di e si y.
Does he sys em allow you o di e en ia e whe he he sou ces belong o aca-
demic ex s, p ess, g ey li e a u e, o o he un egula ed sou ces?
23 Capilla i y
A e he sou ces connec ed a leas a he pa ag aph o sec ion le el?
2.4
Well o med
A e he sou ces p esen ed in a o ma ha is easy o expo , manage, and ci e
sou ces?
3. Sugges ed
p omp s
3.1
Chance
Do he sugges ed new p omp s seem app op ia e o imely gi en he in o ma ion
needs?
3.2 Va ie y
A e he p omp s a ied and help b oaden he ocus o he opic?
4. Idio unc ions 4.1
Speci ic and exclusi e unc ions o each sys em conside ed
In addi ion o he common unc ions examined, does he sys em ha e any o he
speci ic unc ions?
168
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Table 3
Theo e ical sco es.
Pa ame e Code Va iables Theo e ical sco e
Na a i e syn hesis
1.1 Sc a chpad
0-3
1.2 Join
1.3 Connec ion
1.4 Ex ension
1.5 Mul imodali y
Sou ces
2.1 Numbe
2.2 Di e si y
23 Capilla i y
2.4 Well o med
Sugges ed P omp s 3.1 Chance
3.2 Va ie y
Idio unc ions 4.1 Speci ic and exclusi e unc ions o each sys em conside ed 0-3
The sco ing scale is o e ed as an example. Fo each use, hose esponsible may (wi h jus i ica-
ion) de e mine o he scales. In his case, we ha e used a scale ypical o heu is ic e alua ions
in he ield o in o ma ion sys ems usabili y, and i co esponds o he ollowing es ima e:
Punc ua ion In e p e a ion
0 Absence o unc ion o a iable conside ed
1 The unc ion o a iable appea s in a minimal exp ession
2 The unc ion o a iable is co ec ly implemen ed bu allows o imp o emen s
3 The unc ion o a iable is ully implemen ed
The ini ial sco es in his sco ing sys em a e assigned in ui i ely and a e adjus ed as new cases
a e examined o allow compa isons. A inal ally is made once all cases a e examined. I is also
common o wo analys s o assign sco es independen ly, hen he sco es a e compa ed, and
disc epancies a e esol ed by consensus. Howe e , he scale and he speci ic p ocedu e o
assigning sco es can be es ablished o each speci ic p ojec .
Tabla 4
Da a ex ac ion able
Cod. Va iable Punc ua ion
Na a i e syn hesis
1.1 Sc a chpad
1.2 Join
1.3 Connec ion
1.4 Ex ension
1.5 Mul imodali y
Sou ces
2.1 Numbe
2.2 Di e si y
23 Capilla i y
2.4 Well o med
169
C i ical hinking and a i icial in elligence in academia: A quali a i e ma ix analysis p ocedu e o e alua ing AI sys ems
Lluís Codina; Elisenda Aguile a-Co a; Ca los Lopezosa; Pe e F eixa
Digi al communica ion. T ends and good p ac ices
Cod. Va iable Punc ua ion
Addi ional P omp s
3.1 Chance
3.2 Va ie y
Idio unc ions
4.1 Speci ic and exclusi e unc ions o each sys em conside ed
TOTAL
Compa a i e summa y able
Sys em Syn hesis Sou ces P omp s Idio uncion TOTAL
The ables abo e a e common examples o ma ix-based analysis sys ems. Fo each speci ic
p ojec , p ojec manage s can modi y any aspec s as app op ia e.
3.6. O he e alua ion modes
I is clea ha di e en e alua ion me hods can be de eloped. Task-based e alua ion is one
signi ican al e na i e, as in Fon -Julián e al. (2024), in which an e alua ion model is de eloped
ha combines quali a i e and quan i a i e analysis p ocedu es unde speci ic asks and wi h a
g oup o wo o mo e use s as judges who ag ee on hei e alua ions.
Ano he signi ican al e na i e a e he benchma king e alua ion me hods, such as hose
ha can be seen on he A i icial Analysis po al (h ps://a i icialanalysis.ai/) whe e dozens o
language models a e pe iodically compa ed based on a ba e y o es s.
3.7. Di e en ial con ibu ion o his e alua ion mode
Any e alua ion me hod can be use ul, depending on he con ex and objec i es o each case.
The quali a i e e alua ion me hod we p opose he e has a h ee old unc ion:
– S eng hen use s’ c i ical hinking ega ding AI.
– P o ide a me hod o eaching/lea ning and acqui ing skills in he use o AI sys ems.
– P o ide a p ocedu e o e alua ing and compa a i ely analyzing AI sys ems based on
quali a i e ma ix analysis o pa ame e s and indica o s.
3.8. Va iable geome y p ocedu e
The MASIA p ocedu e p o ides analy ical amewo ks ha can be applied as p esen ed.
Howe e , i is no essen ial o use all a iables, and new a iables can be added o e en o he
pa ame e s can be conside ed. The essence o his e alua ion me hod is as ollows:
– The designe s o he analysis, wi h any o he h ee objec i es s a ed abo e, may
conside he con enience o adding new a iables o emo ing some o he a iables.
