The credible Tarot

[eight bits]

Assuming we're finished with the silliness, it seems to me there is at least one serious methodological flaw in the 2004 experiment. The design depends critically on the reader (the first exprimenter) not knowing which spread is which. Otherwise, this information can be communicated to the subject, buried beyond possibility of detection afterwards in the patter of the reading. Such messaging might also discriminate between believer and nonbeliever subjects, as the reader feels "close" or "distant" to or from a subject.

In other words, unless it's truly double blind, then the whole result may have been "dictated" by the reader, whether unconsciously (more likely) or deliberately (less likely).

Is it double blind? Not necessarily. The conclusory sttement is made (p.438),

The cards were then positioned in a way that left no clues for E1 or the participant to be able to distinguish between the layouts.

But how was this achieved? Only the following is stated:

This was achieved by ensuring that all the cards were the same way up and that the separation between cards was approximately equal in the two spreads.

In other words, the experimenters took care that there were no visual clues. What about other sensory modalities?

The method has one spread made by the subject shuffling the cards (p.437). The other is made with minimal handling of the deck by the second experimenter. The methods disclosure doesn't say, but the same physical deck may have been used in each trial. In other words, the deck may smell like the lab. Presumably, the typical subject is not otherwise a regular at the lab, while the second experiementer is a regular. I propose that a human being may be able to discriminate by smell whether objects have been handled by an unfamiliar person rather than a familiar one.

Richard Feynman (Surely you're joking..., chapter "Testing bloodhounds") experimented casually with improving his sense of smell and achieved impressive results with little effort. Some equally informal replication, specifically including trying to determine whether a book had been recently handled by someone, is reported here:

http://www.thescient...re-it-goes.html

There is also some more formal investigation of human scent performance in a tracking task:

http://www.nature.co...abs/nn1819.html

Of course, the peer review policy at Nature may not be up to everybody's standards, lol.

[ChrLzs]

Silliness, eh? I'd call it silliness when someone responds to my comment about the controls not being fully documented, by saying they were 'mentioned' in the very first paragraph. Yeah, 'mentioned' means fully explained, showing examples.... oh wait... no it doesn't, so all of my words were ignored, just as in previous posts.

And I find it hilarious that now the 'silliness' is over... NOW, eight bits, you suddenly DO find one thing wrong...

That's all you found? You missed several elephants.

This time I truly am out. Wake me when the paper appears - 8b, I'm sure now that you spotted that one issue, you can be adviser to a successful study!!

[Tiggs]

Silliness, eh? I'd call it silliness when someone responds to my comment about the controls not being fully documented, by saying they were 'mentioned' in the very first paragraph. Yeah, 'mentioned' means fully explained, showing examples.... oh wait... no it doesn't, so all of my words were ignored, just as in previous posts.

This is what it says:

In this study, the real reading is one for which the participants choose their own cards compared with the control reading, which consists of randomly chosen cards.

Later, in the methodology, it states:

Then both the participant and E1 left the room and a third person (Experimenter 2, E2) came into the room and took the five cards from the bottom of the deck and laid them out in the same pattern as the real reading, only on the left. This spread is the control reading.

So, there you go. Simple, elegant, documented fully - and an integral part of the experiment, given that the entire experiment is based on the participant's subjective assessment of the real reading vs their subjective assessment of the control reading.

And I find it hilarious that now the 'silliness' is over... NOW, eight bits, you suddenly DO find one thing wrong...

That's all you found? You missed several elephants.

Well, do feel free to stick around, and point them out for us. I have an elephant gun, and everything.

[Tiggs]

Assuming we're finished with the silliness, it seems to me there is at least one serious methodological flaw in the 2004 experiment. The design depends critically on the reader (the first exprimenter) not knowing which spread is which.

Agreed - and since we're talking flaws and methodologies:

It's entirely possible that the reader picks up on cues from the participant whilst giving the first reading and adapts his second reading accordingly.

I'm not sure how I feel about the other cards for the reading being taken from the same deck as the first. In doing so, you've reduced the possible card combinations that can appear in that reading (and entirely guaranteed that no two cards will repeat between the two readings).

I'm not sure that telling the participant shuffling the deck that the second reading will be taken from the bottom of the deck is such a great idea, either.

I'm also considering whether or not the questions regarding the Australian Sheep-Goat Scale may be best done upfront, as the results of the experiment may sway the participant, accordingly.

I'm thinking of removing the first experimenter, entirely, by making the experiment computer-based. This is the methodology I'm thinking of:

1. Participant registers age, gender and takes Australian Sheep-Goat Scale assessment.

2. Participant sees Tarot deck on screen, which they can manipulate via a shuffling mechanism of some sort, until they are satisfied.

3. Participant is given two readings of five cards. One from the top of the deck of the Tarot cards they've shuffled (the real reading), one chosen entirely at random by the computer from an entire Tarot deck (our control reading). The reading that they receive first (real or control) will be chosen at random. Within each reading, each card is shown in the same order along with it's common interpretations, in order for the participant to decide for themselves what the cards signify to them.

4. Participant records their feedback on the two readings, using the same questions and scale as the paper we're currently examining. I'm also thinking of asking them an additional question - which reading do they think was the one they picked?

I think that by cutting out experimenter one, you remove a lot of potential for bias from the experiment. Also - given that the tarot cards are digital, you ensure that their backs are actually identical and that the participant will not be able to identify them by any of the accepted senses.

In order to make the randomness as random as I can - I'm planning to use random.org as a generator for the random numbers, which uses atmospherics, rather than a software algorithm.

That's about as far as I've got.

[eight bits]

That's pretty far along.

Getting the machine to provide a good warm and fuzzy reading experience is an overlooked artificial intelligence challenge. Two equally off-putting mechanical readings might be a problem for getting any results at all.

I wasn't too concerned about the human in the loop IF I was more confident about the blindness of the double-blind implementation. I also wasn't concerned that the control reading could not have any cards in common with the trial layout. (Philosophically, I am unsure how the "control" wasn't just as much the result of the subjects' shuffling as the trial... but I am not much of a Tarot theorist.)

It is interesting that the 2004 experimenters didn't ask outright (BOTH the reader and the subject) which spread they thought was which. Since the actual result was "counterintuitive" among the believers, wouldn't it be interestting if they not only gave higher marks to the control reading, but higher marks to what they thought was the control reading? Or, alternatively, nice to know flat out that they guessed wrong, if that was how it was.

[Tiggs]

That's pretty far along.

Getting the machine to provide a good warm and fuzzy reading experience is an overlooked artificial intelligence challenge. Two equally off-putting mechanical readings might be a problem for getting any results at all.

I don't think it's particularly feasible to get a machine to provide a human-style reading. That's why I was planning to emulate the current online tarot reading sites - which just provide the cards and their interpretations - and let the participant do the actual application to their own situation. That way, you also shortcut any issues with the reader's ability to spin a good story / reader fatigue. etc.

Admittedly, it wouldn't be as compelling an experience for the subject, but I guess I'm working on the principle that the less humans you stick in the experiment, the less subject it is to any unconscious bias on their behalf creeping in.

I wasn't too concerned about the human in the loop IF I was more confident about the blindness of the double-blind implementation. I also wasn't concerned that the control reading could not have any cards in common with the trial layout. (Philosophically, I am unsure how the "control" wasn't just as much the result of the subjects' shuffling as the trial... but I am not much of a Tarot theorist.)

I think that the control in that experiment is a result of the shuffling, especially since the subject is told that the cards on the bottom will be used for the second reading - hence why I think that the control should be as random as possible.

In order to remove issues with the double blind implementation, in a human version, the second experimenter could:

1. Duplicate the cards selected using a second deck of cards.

2. Shuffle the remaining cards / a third deck and use that to select five cards with, as the control.

Alternately - to remove any unconscious signalling as to which reading is which in the subsequent layout of the cards by the second experimenter - the cards could instead be represented on a computer display, ensuring perfect positioning (and removing any other potential sensory signalling).

It is interesting that the 2004 experimenters didn't ask outright (BOTH the reader and the subject) which spread they thought was which. Since the actual result was "counterintuitive" among the believers, wouldn't it be interestting if they not only gave higher marks to the control reading, but higher marks to what they thought was the control reading? Or, alternatively, nice to know flat out that they guessed wrong, if that was how it was.

Indeed - I think it's an interesting qualifier to their ratings. Everything being even - it should split around 50/50. Would be interesting if it didn't.

[eight bits]

I wonder if the "warm and fuzzy" effect might be achieved by hijacking the I Ching. Here is one of several "correspondences" proposed:

http://hermenes.com/Homepage/ichintaro_en.htm

You could also easily match a 6-card draw to a selection of a hexagram from the universe of 2^6.

I took a look at the humorously named Australian sheep-goat (what a straight line that is, if only this weren't a family-friendly forum)...

http://aiprinc.org/Thalbournes_ASGS.pdf

That seems straightforward enough for automated self-administration.

Finally,what do you intend to ask participants afterwards about their experiences?

[Tiggs]

I wonder if the "warm and fuzzy" effect might be achieved by hijacking the I Ching. Here is one of several "correspondences" proposed:

http://hermenes.com/...chintaro_en.htm

You could also easily match a 6-card draw to a selection of a hexagram from the universe of 2^6.

I could potentially use that - though I think that the (entirely out of copyright and now in the Public Domain) set of interpretations for the Tarot by Waite aren't actually that bad.

For example - the Eight of Wands via Waite reads:

The card represents motion through the immovable-a flight of wands through an open country; but they draw to the term of their course. That which they signify is at hand; it may be even on the threshold. Divinatory Meanings: Activity in undertakings, the path of such activity, swiftness, as that of an express messenger; great haste, great hope, speed towards an end which promises assured felicity; generally, that which is on the move; also the arrows of love. Reversed: Arrows of jealousy, internal dispute, stingings of conscience, quarrels; and domestic disputes for persons who are married.

Using the I Ching correspondence you've suggested, the Eight of Wands corresponds with 16 - Yu. Legge's set of interpretations (also in the public domain) reads:

Yü indicates that, (in the state which it implies), feudal princes may be set up, and the hosts put in motion, with advantage.

1. The first SIX, divided, shows its subject proclaiming his pleasure and satisfaction. There will be evil.

2. The second SIX, divided, shows one who is firm as a rock. (He sees a thing) without waiting till it has come to pass; with his firm correctness there will be good fortune.

3. The third SIX, divided, shows one looking up (for favours), while he indulges the feeling of pleasure and satisfaction. If he would understand!--If he be late in doing so, there will indeed be occasion for repentance.

4. The fourth NINE, undivided, shows him from whom the harmony and satisfaction come. Great is the success which he obtains. Let him not allow suspicions to enter his mind, and thus friends will gather around him.

5. The fifth six, divided, shows one with a chronic complaint, but who lives on without dying.

6. The topmost six, divided, shows its subject with darkened mind devoted to the pleasure and satisfaction (of the time); but if he change his course even when (it may be considered as) completed, there will be no error.

Perhaps there's a different version of the I Ching readings in the public domain I could take a look at - but I think with a little modernisation, the Waite version should be okay.

I took a look at the humorously named Australian sheep-goat (what a straight line that is, if only this weren't a family-friendly forum)...

http://aiprinc.org/T...ournes_ASGS.pdf

That seems straightforward enough for automated self-administration.

Yes it does. Took me a while to find a copy, and I wasn't entirely sure it was the correct set of questions - so thank you for that.

Finally,what do you intend to ask participants afterwards about their experiences?

From the Paper:

After the readings the participant was asked to rate each of the readings by answering eight questions. The questions asked the participant to say to what extent the reading related to the specific issue s/he had in mind. Three questions were about the reading in general (in terms of interest, insightfulness, and helpfulness) and the other five asked specifically about each card. Answers could range from 1 (not at all) to 8 (very much).

Aside from the extra question - "Which reading do you think was the one you chose?" - I was thinking of asking questions as similar to the original as possible, but I haven't really considered the questions in great detail, to be honest. I'm somewhat tempted to email the original researchers and see if they still have a copy of the exact questions, so I could steal / borrow them in the name of Science.

[eight bits]

Yes, emailing them is a good idea. They're both still around and active. Good luck with this project.